CN114638152B - Deep sea Argo profile buoy energy management method based on HGP-MPC - Google Patents
Deep sea Argo profile buoy energy management method based on HGP-MPC Download PDFInfo
- Publication number
- CN114638152B CN114638152B CN202210148365.3A CN202210148365A CN114638152B CN 114638152 B CN114638152 B CN 114638152B CN 202210148365 A CN202210148365 A CN 202210148365A CN 114638152 B CN114638152 B CN 114638152B
- Authority
- CN
- China
- Prior art keywords
- buoy
- data
- energy management
- prediction
- argo
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000007726 management method Methods 0.000 title claims abstract description 66
- 238000000034 method Methods 0.000 claims abstract description 55
- 238000005457 optimization Methods 0.000 claims abstract description 42
- 230000008569 process Effects 0.000 claims abstract description 35
- 230000007613 environmental effect Effects 0.000 claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 13
- 238000005096 rolling process Methods 0.000 claims abstract description 12
- 238000009826 distribution Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims description 36
- 238000012549 training Methods 0.000 claims description 22
- 238000012360 testing method Methods 0.000 claims description 15
- 239000013535 sea water Substances 0.000 claims description 13
- 230000033001 locomotion Effects 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 10
- 230000005540 biological transmission Effects 0.000 claims description 9
- 238000013480 data collection Methods 0.000 claims description 7
- 230000001133 acceleration Effects 0.000 claims description 6
- 238000007476 Maximum Likelihood Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 230000002068 genetic effect Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000011425 standardization method Methods 0.000 claims description 3
- 230000003190 augmentative effect Effects 0.000 claims description 2
- 238000004140 cleaning Methods 0.000 claims description 2
- 238000005265 energy consumption Methods 0.000 abstract description 12
- 238000011160 research Methods 0.000 abstract description 2
- 238000013139 quantization Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 28
- 230000009471 action Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012821 model calculation Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000013256 coordination polymer Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013501 data transformation Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 229910052744 lithium Inorganic materials 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06313—Resource planning in a project environment
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J7/00—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries
- H02J7/00032—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries characterised by data exchange
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J7/00—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries
- H02J7/0047—Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries with monitoring or indicating devices or circuits
- H02J7/0048—Detection of remaining charge capacity or state of charge [SOC]
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Power Engineering (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Educational Administration (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to a deep sea Argo profile buoy energy management method based on HGP-MPC, which is characterized in that a probability distribution model is established based on HGP, energy consumption estimation of a buoy single profile operation process is carried out on uncertain sea conditions, after accurate prediction of Argo profile buoy required power is completed, a power prediction result is input into an energy management strategy based on MPC, and energy management rolling optimization is carried out on the buoy in a prediction time domain by combining the battery residual electric quantity of the buoy. According to the scheme, the power prediction of the buoy load demand is carried out by adopting a heteroscedastic Gaussian process regression model, energy management research is carried out based on an uncertainty quantization result, the environmental adaptability of an energy management strategy is improved, the buoy power distribution is carried out by adopting a model prediction control method, a prediction time domain is set, rolling optimization is carried out, and real-time decision is realized; the power prediction algorithm is combined with the real-time energy management strategy, so that the environmental adaptability of the energy management strategy is improved, the instantaneity is ensured, and the method has higher practical application and popularization values.
Description
Technical Field
The invention relates to the field of deep sea Argo profile buoy energy management, in particular to a deep sea Argo profile buoy energy management method based on HGP-MPC.
Background
The energy problem is one of the difficulties commonly encountered by underwater vehicles, the deep sea detection working conditions are complex and changeable, and for buoys, the density, temperature and buoy movement speed of sea water at different depths have obvious influence on the energy consumption of the buoys for completing single-section movement. The Argo profile buoy has the characteristics of excellent real-time performance, continuity, high efficiency and the like, and plays an important role in the field of deep sea exploration.
There are a number of factors affecting buoy energy consumption, and the dead time and life of Argo profile buoys are greatly limited by the power system capacity and energy management strategies. The buoy can not actively perform horizontal movement adjustment, and mainly drifts along with ocean currents, so that even if multiple submergence tests are performed in the same sea area, larger uncertainty exists in energy consumption of the buoy in the whole section movement process. In addition, the real-time ocean current parameters of the operating sea area, which are preset by the staff, can lead to uncertainty of energy consumption of buoy requirements. All the factors can influence the estimation accuracy of the buoy for single profile energy consumption, thereby reducing the reliability of the profile cycle times set by the buoy in the life cycle.
In order to better manage the energy of the buoy, after accurately predicting the power required by the Argo profile buoy, an effective energy management strategy needs to be developed in order to reasonably distribute the power of the buoy in the real-time single profile movement process. In the existing energy management strategy, on one hand, the rule-based energy management strategy is difficult to adapt to changeable sea conditions; on the other hand, the energy management strategy based on the optimization algorithm faces the fixed working condition and large calculated amount, and is difficult to meet the real-time application requirement. Therefore, developing an optimized energy management strategy with real-time adaptability is a highly desirable problem.
Disclosure of Invention
The invention provides a deep sea Argo profile buoy energy management method based on HGP-MPC, which aims to solve the defects that the existing energy management strategy is difficult to adapt to changeable sea conditions, is difficult to meet real-time application requirements and the like, and can not only improve the environmental adaptability of the energy management strategy, but also ensure the instantaneity of a prediction algorithm.
The invention is realized by adopting the following technical scheme: the management method of the deep sea Argo profile buoy energy management system based on the HGP-MPC comprises a historical working condition data collection module, a real-time data acquisition module, a terminal processing module, an execution module, a data transmission module and a server processing module; wherein the server processing module comprises an HGP based load power prediction unit and an MPC based energy management unit, characterized in that the method comprises the steps of:
step A, load power prediction based on HGP:
A1, transmitting the obtained historical working condition data to a server processing module through a terminal processing module based on a historical working condition data collecting module;
A2, carrying out power prediction by a load power prediction unit based on HGP;
(1) Feature selection and data preprocessing: determining input and output parameters of the HGP model according to the longitudinal dynamics equation of the Argo profile buoy, and performing data cleaning, data integration, data change and data reduction on the data to improve the data quality; after data preprocessing, randomly dividing a training set and a testing set for the data set;
(2) Determining input characteristics according to a buoy longitudinal dynamics model, constructing a power demand prediction model, selecting a Gaussian process regression kernel function, and determining the number of model hyper-parameters;
the power demand prediction model is as follows:
yi＝f(xi)+εi,xi∈R,yi∈Rd,εi～N(0,σi)
Where x i is the input of the ith sample; y i is the output of the ith sample, i.e., power, all from the training set F () is a regression function; epsilon i is the output noise, obeying a gaussian distribution with a mean value of 0 and a variance of sigma i; the heteroscedastic problem is described as σ i＝r(xi);
(3) Training a power demand prediction model: after the number of the super parameters of the model is determined, an optimization function is established, and the super parameters are optimized;
(4) Verifying the prediction performance of the power demand model, and verifying the trained Argo profile buoy power demand prediction model through the divided test set so as to accurately predict;
Step B, MPC-based energy management optimization
Step B1, based on the real-time data acquisition module, the obtained real-time working condition data are transmitted to a server processing module through a terminal processing module;
step B2, performing energy management optimization by the energy management unit based on the MPC:
(1) Selecting the rotation speed of a driving motor as a control variable according to the longitudinal dynamics equation of the Argo profile buoy;
(2) Setting constraint conditions according to the equipment specifications of the power system and the profile motion conditions of the buoy;
(3) Optimizing an objective function to obtain a control variable signal in a prediction time domain;
inputting a power prediction result into an energy management unit based on MPC, and carrying out energy management optimization on the buoy in a prediction time domain by matching with the battery residual electric quantity acquired by an electric control unit in the buoy;
And C, sending the rotating speed instruction of the driving motor in the prediction time domain obtained in the step B to a terminal processing module through a data transmission module, and sending the rotating speed instruction to a buoy execution module through the terminal processing module to realize real-time rolling optimization.
Further, in the step B2, in the energy management optimization process:
firstly, combining acquired seawater environment parameters, hydraulic system state parameters and battery state parameters; the load power prediction unit based on HGP is utilized to predict the required power in the prediction time domain;
And then optimizing the obtained control action command in the prediction target time domain, namely the driving motor rotating speed command by taking the minimum equivalent energy consumption in the prediction time domain as an objective function.
Finally, sending the instruction to an executing mechanism, and completing single real-time optimization; and after the motor finishes the action within the moment of the designated step length, repeating the process, and performing rolling optimization to obtain a real-time rotating speed instruction of the driving motor until the buoy finishes the single-section operation task.
Further, the real-time working condition data type is consistent with the history working condition data type, the history working condition data comprises environmental parameters collected by the Argo profile buoy after completing the multiple complete submerged floating profile movement process and parameters representing the state of the power system, the environmental parameters comprise sea water temperature, salinity and pressure, and the parameters representing the state of the power system comprise the pressure load of the hydraulic system, the rotating speed of a driving motor, the voltage, the current and the residual electric quantity of a battery.
Further, in the step A2, the input parameters include an initial input including a sea water temperature, a salinity and a pressure, and an augmented input including a velocity, an acceleration and a velocity square calculated based on the initial input; the output parameter includes power.
Further, in the step A2, the following method is mainly adopted when the data preprocessing is performed:
firstly, removing outliers, noise and missing values outside a data 3 sigma range by adopting a Laida criterion;
secondly, smoothing data by adopting an exponential moving average method to reduce errors of the original data caused by measurement errors and sensor hysteresis;
Finally, the original working condition data is standardized by adopting a z-score standardization method, and dimension and magnitude differences among different dimension data are eliminated.
Further, in the step A2, when the power demand model is trained, based on the maximum likelihood estimation method, a negative log marginal likelihood function is established as an optimization function by using training set data, and a genetic algorithm NSGA-II with elite strategy and non-dominant ordering is used for optimizing, so as to obtain an optimal solution of the super parameter.
Compared with the prior art, the invention has the advantages and positive effects that:
According to the scheme, data such as environmental parameters, hydraulic system power parameters and battery state parameters obtained in an Argo profile buoy historical sea test are taken as raw data, and a load power prediction model is established based on a heteroscedastic Gaussian process regression model; then according to the prediction result, energy management optimization is realized on the buoy in a prediction time domain, and a driving motor rotating speed instruction under the minimum energy consumption is output;
The prediction confidence interval of the buoy operation process required power can be provided by adopting the heteroscedastic Gaussian process regression model, so that uncertainty caused by external environment change and internal system dynamic performance can be more accurately described, and the confidence of a prediction result is improved; the uncertain sea conditions and environmental factors are quantitatively modeled through heteroscedastic Gaussian process regression, and energy management research is conducted on the basis of uncertainty quantitative results, so that the environmental adaptability of an energy management strategy can be improved;
The buoy power distribution is carried out by adopting a model predictive control method, a predictive time domain is set, rolling optimization is carried out, and real-time decision is realized; the power prediction algorithm is combined with the real-time energy management strategy, so that the environmental adaptability of the energy management strategy can be improved, and the instantaneity of the prediction algorithm can be ensured.
Drawings
FIG. 1 is a schematic flow diagram of a buoy energy management method according to an embodiment of the present invention;
FIG. 2 is a flowchart of a training process for a heteroscedastic Gaussian process regression model according to an embodiment of the invention.
Detailed Description
In order that the above objects, features and advantages of the invention will be more readily understood, a further description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced otherwise than as described herein, and therefore the present invention is not limited to the specific embodiments disclosed below.
With the development and application of artificial intelligence algorithms such as machine learning, deep learning and the like, when input, output and environmental data of the Argo profile buoy power system under the historical operation condition are obtained, the input, output and environmental data can be used as a training data training model for predicting the required power of the buoy power system under the condition. After reasonably selecting the model and adjusting the parameters, the prediction of the required power of the buoy power system under the working condition can be realized. In order to accurately predict the power demand of the buoy power system and characterize uncertainty of a prediction result, the power demand prediction model of the buoy under different working conditions is established by adopting a heteroscedastic Gaussian process regression (Heteroscedastic Gaussian Process Regression, HGP) model. The Gaussian process regression (Gaussian Process Regression, GPR) model is a regression model based on a Bayesian probability framework, the uncertainty of a predicted variable is quantified through probability reasoning, a predicted mean and variance characteristic uncertainty measure is provided, and a probability model of buoy demand power can be established according to the characteristics of sea water density, pressure, buoy speed, acceleration and the like. Meanwhile, the influence degree of environmental factors on the required power under different working conditions is considered to be different, the different variance model is adopted for evaluation, the influence of the external environment and the internal system state on the buoy is described more accurately, and the difference on the power requirement is shown at different sea depths.
The energy management strategy configured by the buoy needs to meet the requirements of real-time performance, rapidity and effectiveness, and is different from huge calculation amount of a dynamic planning algorithm, the model prediction control method is used for carrying out energy management through real-time online rolling optimization, and has the advantages of low real-time calculation amount and strong disturbance resistance and dynamic correction capability. According to the invention, a model predictive control (Model Predictive Control, MPC) is adopted to build a buoy real-time operation energy management strategy, the required load power in a predicted time domain obtained by a load power prediction unit based on HGP is utilized, an optimization function is built by minimum equivalent energy consumption, and the rotation speed of a driving motor is used as a control target variable to perform rolling optimization.
Specifically, this embodiment proposes a deep sea Argo profile buoy energy management system management method based on HGP-MPC, the energy management system includes a historical working condition data collection module, a real-time data collection module, a terminal processing module and a server processing module, the historical working condition data collection module and the real-time data collection module are both connected with the terminal processing module, the terminal processing module is connected with the Argo buoy through an execution module, the terminal processing module is connected with the server processing module through a data transmission module, wherein the server processing module includes an HGP-based load power prediction unit and an MPC-based energy management unit, and specific implementation steps of the energy management method are as shown in fig. 1, and the method includes:
step A: load power prediction based on HGP:
A1, collecting complete profile data in an Argo buoy sea test based on a historical working condition data collecting module, and transmitting the obtained historical working condition data to a terminal processing module; the terminal processing module sends the history working condition data to the server processing module through the data transmission module, and the load power prediction unit based on the HGP performs data processing and model calculation;
The historical working condition data collection module mainly uses sensor equipment carried by buoy equipment to dynamically acquire environmental parameters, uses a sensor carried by a hydraulic system to acquire hydraulic system state parameters, and uses an electronic control unit in the buoy equipment to acquire battery state parameters, wherein the parameters are all complete profile data acquired in an Argo buoy sea test. Wherein the buoy device refers to a deep sea self-sustaining intelligent Argo profile buoy system; the sensor equipment comprises a temperature and salt depth profiler, a pressure sensor and an electric control system parameter acquisition system. The working condition refers to the combination of the operation state of the buoy power system and environmental data, which are acquired when the buoy performs profile movement, and comprises two stages of submergence and floating, and the corresponding historical working condition data refers to the parameters representing the state of the power system, such as the sea water temperature, salinity, pressure and the like, acquired by the Argo profile buoy during the complete submergence and floating profile movement process, the pressure load of the hydraulic system, the rotation speed of a driving motor, the battery voltage, the current, the residual electric quantity and the like. In this embodiment, the working condition data is mainly derived from a plurality of performance test tests of deep sea Argo buoy developed by the martina sea ditch organization in China.
A2, carrying out power prediction by a load power prediction unit based on HGP;
(1) Performing feature selection according to professional knowledge and machine learning model characteristics, and preprocessing historical data;
when the HGP power prediction model is trained, the input and the output of the HGP model are determined according to the longitudinal dynamics equation of the Argo profile buoy, wherein the initial input is sea water temperature, salinity and pressure, sea water depth information is converted based on the initial input, the speed, the acceleration and the square of the speed are calculated through first-order and second-order difference to serve as the broadening input, the model input consists of the initial input and the broadening input, and the output is the power calculated based on the battery voltage and the current, as shown in the attached table 1.
TABLE 1 input/output selection of heteroscedastic Gaussian process regression model
Data preprocessing refers to data cleansing, data integration, data transformation, and data reduction to improve data quality. The main flow is as follows: firstly, outliers, noise and missing values outside the 3 sigma range of the data are removed by using a Laida criterion; secondly, smoothing data by adopting an exponential moving average method to reduce errors of the original data caused by measurement errors and sensor hysteresis; finally, the original working condition data is standardized by adopting a z-score standardization method, the dimension and magnitude differences among different dimension data are eliminated, and the data form is arranged.
After data preprocessing, randomly dividing a training set and a testing set of the data set, and randomly selecting 80% of the obtained data set as the training set for model training; the remaining 20% was used as a test set to verify model predictive capability.
(2) Constructing a power demand model according to the model input dimension, and selecting a Gaussian process regression kernel function;
When the power demand model is constructed (model initialization), considering that the output distribution corresponding to different inputs is different, the noise variance for fitting the output distribution is also different, and in this embodiment, a different variance gaussian process regression model (HGP) is adopted, and the expression is shown in table 2.
TABLE 2 heteroscedastic Gaussian process regression model mathematical expression
The kernel function selection means that regression fitting is performed on the training set data by using a plurality of kernel functions (including but not limited to the kernel functions shown in the attached table 3), and the kernel function with the best fitting effect is selected as the kernel function used in the gaussian process regression model training unit.
TABLE 3 kernel function type for regression fit
The present embodiment uses a square index kernel as the kernel function used in the gaussian process regression model training unit.
(3) Training a Gaussian process regression model: after the number of the super parameters of the model is determined in the step (2), an optimization function is established, and an intelligent optimization algorithm is adopted to perform super parameter optimization;
After initializing a prediction model form, gaussian process regression model training is performed, the quantity of super-parameters required to be optimized in a kernel function and a noise model is determined, and super-parameter optimization function derivation is performed based on a maximum likelihood estimation method; establishing a negative logarithmic maximum marginal likelihood function related to the super parameter by using the conditional probability of the training set sample, and taking the negative logarithmic maximum marginal likelihood function as an objective function; optimizing the super-parameters by using a genetic algorithm (NSGA-II) with non-dominant ranking of elite strategy to obtain an optimal solution of the super-parameters with the aim of minimizing the objective function; the training process is shown in fig. 2.
(4) Performing model prediction performance verification by using a test set obtained in the model preprocessing stage, so as to accurately predict the required power of the Argo profile buoy;
Test set verification means that the HGP model training is finished, the prediction performance of the model is verified on the test set, and the using indexes comprise precision indexes RMSE and R 2, interval estimation indexes CP and MWP, and the application indexes are shown in an attached table 4.
Table 4 mathematical expression of precision evaluation index and section estimation evaluation index
The model prediction step is to obtain predicted input before real-time rolling optimization, input the predicted input into a trained heteroscedastic Gaussian process regression model, and output calculation is carried out by a joint probability distribution formula in table 5 to obtain the set load power.
TABLE 5 mathematical expression for predictive output calculation of heteroscedastic Gaussian process regression model
After the accurate prediction of the required power of the Argo profile buoy is completed, a power prediction result is input into an energy management unit based on MPC, the energy management and the optimization are carried out on the buoy in a prediction time domain by matching with the battery residual capacity collected by an electric control unit in the buoy, the residual capacity of the buoy is reasonably distributed, and a driving motor rotating speed instruction under the minimum energy consumption is output, and the method is specific:
Step B, energy management optimization based on MPC:
Step B1, acquiring data such as seawater environment parameters, hydraulic system state parameters, battery state parameters and the like in real time based on a real-time data acquisition module, and transmitting the acquired real-time working condition data to a terminal processing module; and the terminal processing module is used for sending the real-time acquired data to the server processing module through the data transmission module, and the energy management unit based on the MPC is used for data processing and model calculation. The real-time data are consistent with the type of the historical working condition data, and refer to environmental parameters such as sea water temperature, salinity, pressure and the like acquired in real time when the buoy performs profile test, hydraulic system load pressure, voltage, current and residual electric quantity of a battery and driving motor rotating speed.
Step B2, performing energy management optimization by the energy management unit based on the MPC:
(1) Selecting the rotation speed of a driving motor as a control variable according to the longitudinal dynamics equation of the Argo profile buoy;
(2) Setting constraint conditions according to the equipment specifications of the power system and the profile motion conditions of the buoy;
Constraint condition setting means that when the energy management optimization based on the MPC is carried out, the optimization result is ensured to meet the practical buoy operation limit and objective condition, and the maximum battery operation current, the battery effective charge state interval, the maximum buoy operation speed, the maximum buoy operation acceleration and the maximum driving motor rotation speed are set. Wherein: considering the safety of battery discharge, the maximum operating current of the battery does not exceed the maximum operating safety current I s of the lithium battery pack; considering a theoretical safe discharge interval of the battery, limiting the effective charge state of the battery to a charge interval [ SOC min,SOCmax ]; considering the stability of buoy operation and the continuity of data acquisition, the maximum operating speed of the buoy is not more than 150% of the international general ideal buoy operating speed (0.1 m/s), and the maximum operating acceleration of the buoy is not more than 0.05m/s 2; considering the running stability of the motor and the buoy, the maximum rotating speed of the driving motor does not exceed 150% of the rated rotating speed of the motor.
(3) Optimally setting an objective function to obtain a control variable signal in a prediction time domain, and sending a control instruction to a data transmission module;
The objective function means that the Argo profile buoy has the minimum equivalent energy consumption in the prediction time domain; optimizing the objective function refers to optimizing the objective function to obtain a control variable signal in a prediction time domain; the control variable refers to a drive motor rotational speed command.
The energy management optimization process is that firstly, data such as sea water environmental parameters, hydraulic system state parameters, battery state parameters and the like are transmitted into an energy optimization unit; then, the load power prediction unit based on HGP is utilized to predict the required power in the prediction time domain; then optimizing the obtained control action instruction in the prediction target time domain, namely the rotation speed instruction of the driving motor, by taking the minimum equivalent energy consumption in the prediction time domain as an objective function; transmitting the instruction to an executing mechanism, and completing single real-time optimization; and after the motor finishes the action within the moment of the designated step length, repeating the process, and performing rolling optimization to obtain a real-time rotating speed instruction of the driving motor until the buoy finishes the single-section operation task.
And C, sending a driving motor rotating speed instruction in a future prediction time domain calculated by the service processing module to the terminal processing module through the data transmission module, and sending the driving motor rotating speed instruction to the buoy execution module through the terminal processing module to realize real-time rolling optimization.
To ensure the real-time adaptability of the energy management strategy, firstly, the required voltage and current estimation of the buoy propulsion process is required for uncertain sea conditions. The complex sea condition changes bring difficulty to voltage and current estimation, and the conventional dynamics model is difficult to meet the prediction requirement, so that the probability distribution model is built based on HGP to realize the estimation. Secondly, to achieve the aim of real-time performance, an optimization algorithm capable of meeting the requirement of rapid operation is designed, and a power distribution result is output in real time according to the torque requirement. The scheme introduces a model predictive control method, sets a predictive time domain, performs rolling optimization, realizes real-time decision, outputs a control variable signal, namely the rotating speed of a driving motor, and realizes timely driving and optimization of the buoy.
In addition, it should be emphasized that the energy management optimization scheme, i.e. the proposed power prediction algorithm and the real-time energy management strategy, described in the present invention are also applicable to other autonomous underwater vehicles.
The present invention is not limited to the above-mentioned embodiments, and any equivalent embodiments which can be changed or modified by the technical content disclosed above can be applied to other fields, but any simple modification, equivalent changes and modification made to the above-mentioned embodiments according to the technical substance of the present invention without departing from the technical content of the present invention still belong to the protection scope of the technical solution of the present invention.
Claims (5)
1. The management method of the deep sea Argo profile buoy energy management system based on the HGP-MPC comprises a historical working condition data collection module, a real-time data acquisition module, a terminal processing module, an execution module, a data transmission module and a server processing module; wherein the server processing module comprises an HGP based load power prediction unit and an MPC based energy management unit, characterized in that the method comprises the steps of:
step A, load power prediction based on HGP:
A1, transmitting the obtained historical working condition data to a server processing module through a terminal processing module based on a historical working condition data collecting module;
A2, carrying out power prediction by a load power prediction unit based on HGP;
(1) Feature selection and data preprocessing: determining input and output parameters of the HGP model according to the longitudinal dynamics equation of the Argo profile buoy, and performing data cleaning, data integration, data change and data reduction on the data to improve the data quality; after data preprocessing, randomly dividing a training set and a testing set for the data set;
(2) Determining input characteristics according to a buoy longitudinal dynamics model, constructing a power demand prediction model, selecting a Gaussian process regression kernel function, and determining the number of model hyper-parameters;
the power demand prediction model is as follows:
yi＝f(xi)+εi,xi∈R,yi∈Rd,εi～N(0,σi)
Where x i is the input of the ith sample; y i is the output of the ith sample, i.e., power, all from the training set F () is a regression function; epsilon i is the output noise, obeying a gaussian distribution with a mean value of 0 and a variance of sigma i; the heteroscedastic problem is described as σ i＝r(xi);
(3) Training a power demand prediction model: after the number of the super parameters of the model is determined, an optimization function is established, and the super parameters are optimized;
(4) Verifying the prediction performance of the power demand model, and verifying the trained Argo profile buoy power demand prediction model through the divided test set so as to accurately predict;
Step B, MPC-based energy management optimization
Step B1, based on the real-time data acquisition module, the obtained real-time working condition data are transmitted to a server processing module through a terminal processing module;
step B2, performing energy management optimization by the energy management unit based on the MPC:
(1) Selecting the rotation speed of a driving motor as a control variable according to the longitudinal dynamics equation of the Argo profile buoy;
(2) Setting constraint conditions according to the equipment specifications of the power system and the profile motion conditions of the buoy;
(3) Optimizing an objective function to obtain a control variable signal in a prediction time domain;
inputting a power prediction result into an energy management unit based on MPC, and carrying out energy management optimization on the buoy in a prediction time domain by matching with the battery residual electric quantity acquired by an electric control unit in the buoy;
And C, sending the rotating speed instruction of the driving motor in the prediction time domain obtained in the step B to a terminal processing module through a data transmission module, and sending the rotating speed instruction to a buoy execution module through the terminal processing module to realize real-time rolling optimization.
2. The method of managing a deep sea Argo profile buoy energy management system based on HGP-MPC according to claim 1, wherein: the real-time working condition data type is consistent with the history working condition data type, the history working condition data comprises environmental parameters collected by the Argo profile buoy in the process of completing multiple complete submerged floating profile movement and parameters representing the state of a power system, the environmental parameters comprise the temperature, the salinity and the pressure of sea water, and the parameters representing the state of the power system comprise the pressure load of a hydraulic system, the rotating speed of a driving motor, the voltage, the current and the residual electric quantity of a battery.
3. The method of managing a deep sea Argo profile buoy energy management system based on HGP-MPC according to claim 2, wherein: in the step A2, the input parameters include an initial input including sea water temperature, salinity and pressure and an augmented input including a velocity, acceleration and velocity square calculated based on the initial input; the output parameter includes power.
4. The method of managing a deep sea Argo profile buoy energy management system based on HGP-MPC according to claim 1, wherein: in the step A2, when data preprocessing is performed, the following manner is mainly adopted:
firstly, removing outliers, noise and missing values outside a data 3 sigma range by adopting a Laida criterion;
secondly, smoothing data by adopting an exponential moving average method to reduce errors of the original data caused by measurement errors and sensor hysteresis;
Finally, the original working condition data is standardized by adopting a z-score standardization method, and dimension and magnitude differences among different dimension data are eliminated.
5. The method of managing a deep sea Argo profile buoy energy management system based on HGP-MPC according to claim 1, wherein: in the step A2, when the power demand model is trained, a negative logarithmic marginal likelihood function is established as an optimization function by utilizing training set data based on a maximum likelihood estimation method, and a genetic algorithm NSGA-II with elite strategy and non-dominant ordering is used for optimizing, so that an optimal solution of the super-parameters is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210148365.3A CN114638152B (en) | 2022-02-18 | 2022-02-18 | Deep sea Argo profile buoy energy management method based on HGP-MPC |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210148365.3A CN114638152B (en) | 2022-02-18 | 2022-02-18 | Deep sea Argo profile buoy energy management method based on HGP-MPC |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114638152A CN114638152A (en) | 2022-06-17 |
| CN114638152B true CN114638152B (en) | 2024-05-17 |
Family
ID=81946520
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210148365.3A Active CN114638152B (en) | 2022-02-18 | 2022-02-18 | Deep sea Argo profile buoy energy management method based on HGP-MPC |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114638152B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101911756B1 (en) * | 2018-05-04 | 2018-10-25 | (주)에디넷 | The system for real-time remote monitoring buoys on the sea |
| CN109572937A (en) * | 2018-12-18 | 2019-04-05 | 中国船舶重工集团公司第七0研究所 | A kind of deep-sea control buoy |
| CN112288191A (en) * | 2020-11-19 | 2021-01-29 | 国家海洋信息中心 | Ocean buoy service life prediction method based on multi-class machine learning method |
| CN113378491A (en) * | 2021-05-31 | 2021-09-10 | 青岛海洋地质研究所 | Deep sea Argo buoy energy consumption parameter sensitivity analysis method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8195395B2 (en) * | 2009-09-06 | 2012-06-05 | The United States Of America As Represented By The Secretary Of Commerce | System for monitoring, determining, and reporting directional spectra of ocean surface waves in near real-time from a moored buoy |
-
2022
- 2022-02-18 CN CN202210148365.3A patent/CN114638152B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101911756B1 (en) * | 2018-05-04 | 2018-10-25 | (주)에디넷 | The system for real-time remote monitoring buoys on the sea |
| CN109572937A (en) * | 2018-12-18 | 2019-04-05 | 中国船舶重工集团公司第七0研究所 | A kind of deep-sea control buoy |
| CN112288191A (en) * | 2020-11-19 | 2021-01-29 | 国家海洋信息中心 | Ocean buoy service life prediction method based on multi-class machine learning method |
| CN113378491A (en) * | 2021-05-31 | 2021-09-10 | 青岛海洋地质研究所 | Deep sea Argo buoy energy consumption parameter sensitivity analysis method |
Non-Patent Citations (1)
| Title |
|---|
| 基于机器学习的海洋浮标寿命及轨迹预测;陈若冰;《海洋通报》;20210630;第1-12页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114638152A (en) | 2022-06-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112629863B (en) | Bearing fault diagnosis method for dynamic joint distribution alignment network under variable working conditions | |
| CN109102005B (en) | Small sample deep learning method based on shallow model knowledge migration | |
| CN110221225B (en) | Spacecraft lithium ion battery cycle life prediction method | |
| Fan et al. | A novel machine learning method based approach for Li-ion battery prognostic and health management | |
| CN107944648B (en) | Large ship speed and oil consumption rate prediction method | |
| CN112986827A (en) | Fuel cell residual life prediction method based on deep learning | |
| CN112884008B (en) | Prediction evaluation method and device for running state of power information acquisition system | |
| CN114218872B (en) | DBN-LSTM semi-supervised joint model-based residual service life prediction method | |
| CN117273440B (en) | Engineering construction Internet of things monitoring and managing system and method based on deep learning | |
| CN111190349A (en) | Method, system and medium for monitoring state and diagnosing fault of ship engine room equipment | |
| CN108961460B (en) | Fault prediction method and device based on sparse ESGP (Enterprise service gateway) and multi-objective optimization | |
| CN113313139A (en) | Wind power prediction uncertainty quantification method based on dynamic characteristics of unit | |
| WO2023065580A1 (en) | Fault diagnosis method and apparatus for gearbox of wind turbine generator set | |
| CN111400964B (en) | Fault occurrence time prediction method and device | |
| CN117521512A (en) | Bearing residual service life prediction method based on multi-scale Bayesian convolution transducer model | |
| CN114462305B (en) | Method for optimizing long-endurance energy consumption and estimating service life of Argo buoy | |
| CN114638152B (en) | Deep sea Argo profile buoy energy management method based on HGP-MPC | |
| CN107527093B (en) | Wind turbine generator running state diagnosis method and device | |
| CN117648557A (en) | SOH prediction method and device based on SOH combined noise reduction | |
| CN117664558A (en) | Generator gear box abnormality detection method, device, equipment and storage medium | |
| CN112014757A (en) | Battery SOH estimation method integrating capacity increment analysis and genetic wavelet neural network | |
| CN116819372A (en) | Method for calculating health state of energy storage battery by neural network | |
| CN115456032A (en) | Residual service life prediction method based on deep hybrid multitask model | |
| CN113269356B (en) | Missing data-oriented equipment residual life prediction method and system | |
| Pan | Corrosion detection method of substation Aboveground Steel Structure Based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |