A method and system for transient temperature control of underwater ship water heaters

Abstract

This invention discloses a method and system for transient temperature control of an underwater ship water heater. The method includes: acquiring temperature values ​​T at n monitoring points of the underwater ship water heater. current The corresponding target temperature value T target and real-time power load P ship Based on the temperature value T at the monitoring point current The corresponding target temperature value T target and real-time power load P ship Construct the first state vector s t Based on the first state vector s t Define the heating power adjustment amount ΔQ as action a t Based on the action a t The invention constructs a first piecewise reward function that couples temperature error with energy consumption; based on the above steps, it builds an Actor-Critic network incorporating a spatiotemporal attention mechanism; and it quantizes and deploys the trained Actor network to an embedded controller. This invention offers advantages such as high-precision temperature field control, dynamic load adaptation, improved energy efficiency, system-level energy efficiency optimization, and strong engineering applicability.

Description

Technical Field

[0001] This invention belongs to the field of computer systems engineering, and in particular relates to a method and system for transient temperature control of underwater ship water heaters. Background Technology

[0002] In the field of underwater vessel water heater control, traditional temperature control methods mainly rely on PID (Proportional-Integral-Derivative) control or rule-based logic control. While these methods are simple and easy to implement, they have significant technical limitations in the dynamic and complex underwater environment.

[0003] Underwater ship water heaters typically feature large capacity and multiple heating zones. Traditional single-point temperature measurement combined with PID control cannot detect the spatial non-uniformity of the temperature field, leading to localized overheating or underheating. Experimental data shows that in systems using PID control, the standard deviation of the temperature field can reach ±5℃ or more, significantly affecting the equipment's lifespan.

[0004] Shipboard electrical systems need to supply power to multiple loads, including propulsion, lighting, and communications, and the electrical load P_ship may experience instantaneous fluctuations of more than 20%. Traditional fixed-threshold power allocation strategies often lead to two problems:

[0005] When the load suddenly increases, the heating power is not reduced in time, causing the power system to trip due to overload;

[0006] When the load suddenly drops, the heating power fails to increase quickly, resulting in a delay in temperature recovery (the measured delay is 3-5 minutes).

[0007] Existing methods neglect the nonlinear characteristics of thermodynamic processes:

[0008] The transfer function of heating power and temperature change varies with water flow velocity and ambient pressure;

[0009] Thermal inertia causes a lag in temperature response, making it difficult to adaptively adjust PID parameters. Tests on a certain type of deep-sea work vessel showed that the overshoot of traditional methods exceeded 8°C under varying operating conditions.

[0010] Ship energy systems require strict energy efficiency management, but traditional methods only use temperature deviation as a control target, neglecting energy consumption indicators. Statistics show that existing systems consume as much as 15%-20% of their power ineffectively during steady-state operation.

[0011] The underwater environment places extremely high demands on equipment reliability, but traditional control methods lack the ability to handle temperature gradients. Real-time monitoring may lead to thermal stress cracking of the sealing material (in one case). (This can lead to flange leakage).

[0012] In recent years, although some studies have attempted to introduce fuzzy control and model predictive control (MPC) into this field, the following problems still exist:

[0013] Fuzzy rules rely on expert experience and are difficult to cover all working conditions;

[0014] MPC requires a high-precision thermodynamic model, while underwater environmental parameters (such as seawater permeability) are difficult to model accurately in real time.

[0015] Existing methods have not achieved multi-objective coordinated optimization of load, temperature, and energy consumption. Summary of the Invention

[0016] To address the shortcomings of the existing technology, this invention provides a method for transient temperature control of an underwater ship water heater, comprising the following steps:

[0017] Step 1: Obtain the temperature values ​​T at n monitoring points of the underwater ship water heater. current The corresponding target temperature value T target and real-time power load P ship ;

[0018] Step 2: Based on the temperature value T at the monitoring point current The corresponding target temperature value T target and real-time power load P ship Construct the first state vector s t ;

[0019] Step 3: Based on the first state vector s t Define the heating power adjustment amount ΔQ as action a t ;

[0020] Step 4: Based on the aforementioned action a t Construct the first piecewise reward function that couples temperature error with energy consumption;

[0021] Step 5: Based on the above steps, construct an Actor-Critic network that includes a spatiotemporal attention mechanism;

[0022] Step 6: Quantize and deploy the trained Actor network to the embedded controller.

[0023] Among them, the temperature values ​​T of the n monitoring points in step one current The constructed n-dimensional spatial distribution features are used for the action amplitude constraint in step three and the temperature gradient penalty term calculation in step four.

[0024] In step three, the heating power adjustment amount ΔQ is the action a. t The range of ΔQ is [-f(P)]. ship ),f(P ship)], where f(P ship )=Q max *(1-P ship / P rated P is the load adaptive limiting function. rated Q represents the rated load power of the ship's electrical system. max This indicates the maximum permissible rate of power change of the heater under rated load.

[0025] Among them, the load adaptive limiting function f(P) in step three ship This is determined through the following process:

[0026] Establish a ship load-heating power constraint database, including different real-time electrical loads P ship The maximum allowable heating power Q allow ;

[0027] Q is fitted using cubic spline interpolation allow With P ship Relationship curve;

[0028] Introducing a safety margin coefficient η∈(0,1), we finally determine f(P) ship )=η·Q allow (P ship ).

[0029] In step four, the first segment reward function is calculated using the following formula: Where λ is the dynamic weighting coefficient.

[0030] The spatiotemporal attention mechanism in step five includes:

[0031] The spatial attention layer calculates the attention weight of each temperature measurement point using the following formula. Where i represents the i-th temperature measurement point, v represents a trainable parameter vector used to map hidden features to scalar attention scores, and W represents a trainable weight matrix used to fuse current temperature and target temperature features. T represents the current temperature value at the i-th temperature measurement point; target Indicates the global target temperature value;

[0032] Temporal convolutional layers employ dilated convolutions to extract multi-scale temporal features;

[0033] An action generation layer that outputs Gaussian distributed actions with exploratory noise.

[0034] In step five, the Actor network takes the first state vector from step two as input and outputs the action 'a' from step three. t , with the action a tThe range of the relevant exploration noise is dynamically adjusted according to the cumulative reward calculated based on the first piecewise reward function in step four.

[0035] In step five, the Critic network uses the first state vector from step two and the action a from step three. t The input is the value function, which is used to evaluate the first segmented reward function in step four.

[0036] Specifically, step six includes: performing the action a from step three with a fixed period Δt. t The process is generated, where Δt is adaptively adjusted according to the rate of change of the real-time electricity load described in step one.

[0037] This invention also proposes a transient temperature control system for an underwater ship water heater, comprising:

[0038] The sensor module is used to acquire the temperature values ​​T at n monitoring points of the underwater ship water heater. current The corresponding target temperature value T target and real-time power load P ship ;

[0039] The state vector construction module is used to construct a state vector based on the temperature value T of the monitoring point. current The corresponding target temperature value T target and real-time power load P ship Construct the first state vector s t ;

[0040] The space definition module is used to define the space based on the first state vector s. t Define the heating power adjustment amount ΔQ as action a t ;

[0041] The reward function construction module is used to construct the reward function based on the action a. t Construct the first piecewise reward function that couples temperature error with energy consumption;

[0042] The network building module is used to construct an Actor-Critic network containing a spatiotemporal attention mechanism based on the above steps.

[0043] An embedded controller is used to quantize and deploy the trained Actor network.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] High-precision temperature field control; the spatiotemporal attention mechanism can capture the spatial correlation of n temperature measurement points, reducing the standard deviation of the temperature field to within ±1.2℃ (measured data);

[0046] Dynamic load adaptation, the load adaptive limiting function f(P_ship) achieves millisecond-level power adjustment during load abrupt changes:

[0047] The power response delay is reduced from 3s in the traditional method to 0.1s;

[0048] The power system stability index RI has been improved to 0.92 (out of 1).

[0049] The dynamic weighting coefficient λ enables real-time balance between energy consumption and temperature control, improving energy efficiency by 18%.

[0050] exist Power reduction is triggered within 2ms when the time limit is exceeded, reducing the failure rate to 0.01 times / thousand hours.

[0051] System-level energy efficiency optimization, the reward function λP_ship(ΔQ) 2 The project effectively suppressed ineffective power consumption, improving the overall energy efficiency ratio (COP) from 2.1 to 2.8; it also maintained control continuity and achieved 99.99% availability even in the event of sensor failure.

[0052] With enhanced engineering applicability, the technical effects of this invention have been verified through sea trials. Compared with traditional PID control, temperature control accuracy is improved by 76%, energy consumption is reduced by 22%, and emergency response speed is increased by 40 times, comprehensively solving the technical problems in the background technology. Attached Figure Description

[0053] The above and other objects, features, and advantages of this disclosure will become readily apparent from the following detailed description of exemplary embodiments, taken in conjunction with the accompanying drawings. In the drawings, several embodiments of this disclosure are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0054] Figure 1 This is a flowchart illustrating a transient temperature control method for an underwater ship water heater according to an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram illustrating the parameter logic relationship of steps two to five according to an embodiment of the present invention; and

[0056] Figure 3 This is a network architecture diagram of a Critic network according to an embodiment of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0058] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0059] It should be understood that although the terms first, second, third, etc., may be used to describe... in the embodiments of the present invention, these... should not be limited to these terms. These terms are only used to distinguish... For example, first... may also be referred to as second... without departing from the scope of the embodiments of the present invention, and similarly, second... may also be referred to as first...

[0060] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0061] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0062] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0063] The optional embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0064] Example 1

[0065] like Figure 1 As shown, this invention discloses a transient temperature control method for an underwater ship water heater, comprising the following steps:

[0066] Step 1: Obtain the temperature values ​​T at n monitoring points of the underwater ship water heater. current The corresponding target temperature value T target and real-time power load P ship ;

[0067] Step 2: Based on the temperature value T at the monitoring point current The corresponding target temperature value T target and real-time power load P ship Construct the first state vector s t ;

[0068] Step 3: Based on the first state vector s t Define the heating power adjustment amount ΔQ as action a t ;

[0069] Step 4: Based on the aforementioned action a t Construct the first piecewise reward function that couples temperature error with energy consumption;

[0070] Step 5: Based on the above steps, construct an Actor-Critic network that includes a spatiotemporal attention mechanism;

[0071] Step 6: Quantize and deploy the trained Actor network to the embedded controller.

[0072] Example 2

[0073] This invention proposes a method for transient temperature control of underwater ship water heaters, comprising the following steps:

[0074] Step 1: Obtain the temperature values ​​T at n monitoring points of the underwater ship water heater. current The corresponding target temperature value T target and real-time power load P ship ;

[0075] Step 2: Based on the temperature value T at the monitoring point current The corresponding target temperature value T target and real-time power load P ship Construct the first state vector s t ;

[0076] Step 3: Based on the first state vector s t Define the heating power adjustment amount ΔQ as action a t ;

[0077] Step 4: Based on the aforementioned action a t Construct the first piecewise reward function that couples temperature error with energy consumption;

[0078] Step 5: Based on the above steps, construct an Actor-Critic network that includes a spatiotemporal attention mechanism;

[0079] Step 6: Quantize and deploy the trained Actor network to the embedded controller.

[0080] Among them, the temperature values ​​T of the n monitoring points in step one current The constructed n-dimensional spatial distribution features are used for the action amplitude constraint in step three and the temperature gradient penalty term calculation in step four.

[0081] Among them, T current T target and P ship Composition of state vector s t =[T current ,T target ,P ship ], T current This is the normalized n-dimensional temperature field vector.

[0082] In step three, the heating power adjustment amount ΔQ is the action a. t The range of ΔQ is [-f(P)]. ship ),f(P ship )], where f(P ship )=Q base *(1-P ship / P rated P is the load adaptive limiting function. rated Q represents the rated load power of the ship's electrical system. base This indicates the maximum adjustment amount of the heater under rated load.

[0083] Wherein, the action a t The standardized strategy output is converted into the actual power adjustment amount through a linear mapping, i.e., ΔQ = a t ·f(P ship ).

[0084] Among them, the load adaptive limiting function f(P) in step three ship This is determined through the following process:

[0085] Establish a ship load-heating power constraint database, including different real-time electrical loads P ship The maximum allowable heating power Q allow ;

[0086] Q is fitted using cubic spline interpolation allow With P ship Relationship curve;

[0087] Introducing a safety margin coefficient η∈(0,1), we finally determine f(P) ship )=η·Q allow (P ship ).

[0088] Among them, Q allow (P ship This indicates that the ship is under its current load P. ship The theoretical maximum allowable heating power (unit: kW) is the upper limit of heating power that ensures the stable operation of the ship's electrical system.

[0089] Q allow (P ship ) = min(P rated -P ship Q thermal ), where P rated -P ship Q represents the remaining capacity of the power system. thermal This indicates the thermodynamic limit power of the water heater (determined by its heat dissipation capacity).

[0090] To prevent overload risks caused by load fluctuations and measurement errors, η is set to (0.7, 0.95) (empirical engineering value). η0 represents the basic safety factor (default 0.9), and k represents the load change rate sensitivity coefficient. This represents the absolute value of the instantaneous rate of change of load.

[0091] like Figure 2 The diagram shown is a schematic representation of the parameter logic relationship in steps two through five.

[0092] In step four, the first segment reward function is calculated using the following formula: Where λ is the dynamic weighting coefficient.

[0093] Among them, λ=λ0·xxp(-μt)+λ min t is the number of training steps, μ is the decay coefficient, and λ0 and λ min These are the initial value and the lower limit, respectively.

[0094] The spatiotemporal attention mechanism in step five includes:

[0095] The spatial attention layer calculates the attention weight of each temperature measurement point using the following formula. Where i represents the i-th temperature measurement point, v represents a trainable parameter vector used to map hidden features to scalar attention scores, and W represents a trainable weight matrix used to fuse current temperature and target temperature features. T represents the current temperature value at the i-th temperature measurement point; target Indicates the global target temperature value;

[0096] Temporal convolutional layers employ dilated convolutions to extract multi-scale temporal features;

[0097] An action generation layer that outputs Gaussian distributed actions with exploratory noise.

[0098] Where, v∈R d×1 For trainable projection vectors, W∈R d×2 For trainable weight matrix, With T target The splicing operation achieves local-global feature fusion.

[0099] In step five, the Actor network takes the first state vector from step two as input and outputs the action ΔQ from step three, which is related to the action a. t The range of the relevant exploration noise is dynamically adjusted according to the cumulative reward calculated based on the first piecewise reward function in step four.

[0100] In simple terms, the Actor network outputs action a. t =π θ (s t )+ε, Critic network output state-action value function ε represents the exploratory noise, ε ~ N(0,σ) 2 ), π θ For policy networks.

[0101] The final power command must be satisfied, Q t+1 =Q t +ΔQ∈[0, Q max ],(Q max (This refers to the absolute maximum power of the heater).

[0102] In step five, the Critic network uses the first state vector s from step two. t and action a in step three. t As input, the output value function is used to evaluate the first segmented reward function in step four. For example... Figure 3 The diagram shows the network architecture of the Critic network.

[0103] The mathematical expression of the above content is as follows:

[0104] Specifically, step six includes: generating the action ΔQ of step three at a fixed period Δt, wherein Δt is adaptively adjusted according to the rate of change of the real-time electricity load described in step one.

[0105] Example 3

[0106] This invention also proposes a transient temperature control system for an underwater ship water heater, comprising:

[0107] The sensor module is used to acquire the temperature values ​​T at n monitoring points of the underwater ship water heater. current The corresponding target temperature value T target and real-time power load P ship ;

[0108] The state vector construction module is used to construct a state vector based on the temperature value T of the monitoring point. current The corresponding target temperature value T target and real-time power load P ship Construct the first state vector s t ;

[0109] The space definition module is used to define the space based on the first state vector s. t Define the heating power adjustment amount ΔQ as action a t ;

[0110] The reward function construction module is used to construct the reward function based on the action a. t Construct the first piecewise reward function that couples temperature error with energy consumption;

[0111] The network building module is used to construct an Actor-Critic network containing a spatiotemporal attention mechanism based on the above steps.

[0112] An embedded controller is used to quantize and deploy the trained Actor network.

[0113] The system also includes an anomaly handling module, which is used to estimate missing temperature data using Kalman filtering when a sensor failure is detected; to switch to a local cache control strategy when communication is interrupted; and to force entry into a low-power mode when the energy system is abnormal.

[0114] Example 4

[0115] This disclosure provides a non-volatile computer storage medium storing computer-executable instructions that can perform the steps described in the above embodiments.

[0116] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0117] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0118] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (AN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0119] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0120] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0121] The preferred embodiments of the present invention have been described above to make the spirit of the present invention clearer and easier to understand, and are not intended to limit the present invention. All modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope summarized by the appended claims.

Claims

1. A method of transient temperature control for an underwater vessel water heater, the method comprising: The method comprises the following steps: Step one, obtain the temperature value T of n monitoring points of the underwater ship water heater current , the corresponding target temperature value T target , and the real-time power load P ship ; Step two, constructing a first state vector s t based on the temperature value T current at the monitoring point target , the corresponding target temperature value T target and the real-time power load P ship . Step 3: Based on the first state vector s t Define the heating power adjustment amount ΔQ as action a t ; Step four, based on said action a t a first piecewise reward function coupling temperature error and energy consumption is constructed; Step five, based on the above steps, an Actor-Critic network containing a spatio-temporal attention mechanism is constructed; Step six, the trained Actor network is quantitatively deployed to an embedded controller; The spatio-temporal attention mechanism of the step five comprises: a spatial attention layer that computes attention weights for each temperature measurement point using the following equation where i denotes the i-th temperature measurement point, v denotes a trainable parameter vector for mapping hidden features to scalar attention scores, and W denotes a trainable weight matrix for fusing current temperature and target temperature features, denotes the current temperature value of the i-th temperature measurement point; T target denotes the global target temperature value; a time convolution layer which extracts multi-scale time sequence features by using a cavity convolution; an action generation layer which outputs a Gaussian distribution action with exploration noise; In step five, the Actor network uses the first state vector s from step two. t Given the input, output the action 'a' in step three. t , with the action a t The range of the relevant exploration noise is dynamically adjusted according to the cumulative reward calculated based on the first piecewise reward function in step four; The Critic network in the step five takes the first state vector in the step two and the action a in the step three as input, and outputs a value function V(a | s) for evaluating the first segment reward function in the step four. t The Critic network in the step five takes the first state vector in the step two and the action a in the step three as input, and outputs a value function V(a | s) for evaluating the first segment reward function in the step four.

2. The method of claim 1, wherein, The temperature values T of the n monitoring points in step one current The constructed n-dimensional space distribution features are used for the action magnitude constraint in step three and the temperature gradient penalty term calculation in step four.

3. The method of claim 1, wherein, The heating power adjustment amount ΔQ in step three is action a t , ΔQ is in the range [-f(P ship ), f(P ship )], wherein is a load adaptive clipping function, P rated represents the rated load power of the ship power system, and Q max represents the maximum allowable power change rate of the heater under the rated load.

4. The method of claim 3, wherein, The step three load adaptive clipping function f(P ship ) is determined by the following procedure: A ship load-heating power constraint database is established, containing the maximum allowed heating power Q ship under different real-time power consumption loads P allow ; The Q is fitted by cubic spline interpolation allow The relationship curve between P ship and Q A safety margin coefficient η ∈ (0, 1) is introduced, and finally f(P ship )=η·Q allow (P ship ) is determined.

5. The method of claim 1, wherein, The first segment reward function in the fourth step is calculated by the following formula: where λ is a dynamic weight coefficient, t is the training step number, μ is a decay coefficient, λ0and λ min are the initial value and lower limit, respectively.

6. The method of claim 1, wherein, The step six specifically comprises: performing the action a of step three with a fixed period Δt t generating, wherein Δt is adaptively adjusted according to the rate of change of the real-time power consumption load of step one.

7. An underwater ship water heater transient temperature control system using the method of any one of claims 1-6, comprising: a sensor module for acquiring temperature values T of n monitoring points of the underwater ship water heater current , a respective target temperature value T target and a real-time power load P ship ; a state vector construction module for constructing a first state vector s current based on temperature values T target and real-time power consumption loads P ship of the monitoring points t ; a space definition module for defining a heating power adjustment amount AQ as an action a t based on the first state vector s t ; a reward function construction module configured to construct a first piecewise reward function coupling temperature error and energy consumption based on the action a t , construct a first piecewise reward function coupling temperature error and energy consumption; a network construction module which constructs an Actor-Critic network containing a spatio-temporal attention mechanism; an embedded controller which is used for quantitatively deploying the trained Actor network therein.

Images