Rectification equipment processing method and device and server

Abstract

The invention provides a rectification equipment processing method and device and a server, and relates to an energy management technology, and the method comprises the steps: obtaining first feature parameters of a plurality of rectification equipment in a current period according to a preset recommendation period in a monitoring process of the plurality of rectification equipment; determining a second characteristic parameter according to the first characteristic parameter; and training and updating a preset sleep energy-saving model according to the first characteristic parameter, the second characteristic parameter and training data of a previous period before the current period. According to the first characteristic parameter, the second characteristic parameter and a sleep energy-saving model, performing action prediction on the plurality of rectification devices in the next recommendation period to obtain target recommendation action information; and according to the target recommended action information, changing the use states of the plurality of rectification devices. According to the method, the technical problem that the energy-saving effect is poor due to the fact that energy-saving control of the sleep energy-saving technology of the rectifying equipment is not fine enough is solved.

Description

Technical Field

[0001] The present application relates to energy management technology, and in particular to a rectifier equipment processing method, device and server. Background Art

[0002] The rectifier equipment dormancy energy-saving technology has been widely used in sites. Its basic principle is to achieve system energy saving by switching the rectifier equipment so that the rectifier equipment works in a high-efficiency range. Specifically, in the current mainstream solution, the high-efficiency range of the rectifier equipment is defined by pre-set soft-opening point and soft-closing point. When the load rate of the rectifier equipment is higher than the soft-opening point, the monitoring unit turns on one or more rectifier equipment; when the load rate of the rectifier equipment is lower than the soft-closing point, the monitoring unit turns off one or more rectifier equipment. Through this control method, the load rate of the rectifier equipment is in a relatively high-efficiency range.

[0003] However, in actual applications, the soft-off and soft-on points are set manually, and the default soft-off point is usually 90% and the soft-on point is 40%. The range is wide, and it is difficult to ensure customized adjustments for different rectifier equipment characteristics, resulting in difficulty in implementing refined energy-saving control strategies, which in turn leads to poor energy-saving effects. Summary of the invention

[0004] The present application provides a rectifier device processing method, device and server, which are used to solve the technical problem that the energy-saving control of the rectifier device sleep energy-saving technology is not precise enough, resulting in poor energy-saving effect.

[0005] In a first aspect, the present application provides a rectification device processing method, comprising:

[0006] In the process of monitoring the plurality of rectifying devices, obtaining first characteristic parameters of the plurality of rectifying devices in a current cycle according to a preset recommended cycle;

[0007] According to the first characteristic parameter, a second characteristic parameter is determined; and according to the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle, a preset sleep energy-saving model is trained and updated; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict the target recommended action information according to multiple sleep actions in the sleep action set;

[0008] According to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model, the action of multiple rectifying devices in the next recommended cycle is predicted to obtain target recommended action information; and according to the target recommended action information, the use status of multiple rectifying devices is changed.

[0009] Further, the training and updating of the preset dormancy energy-saving model according to the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle includes:

[0010] Based on the first characteristic parameter and the second characteristic parameter, determine and update the current constraint target value and decision score;

[0011] According to the first characteristic parameter, the second characteristic parameter, the current constraint target value and the training data of the previous cycle before the current cycle, the constraint model in the preset hibernation energy-saving model is trained and updated; and according to the first characteristic parameter, the second characteristic parameter, the decision score and the training data of the previous cycle before the current cycle, the optimization model in the preset hibernation energy-saving model is trained and updated.

[0012] Furthermore, the constraint model includes constraint condition information, and the constraint condition information is:

[0013] |BusU-TarU| <TarUThr,

[0014] Wherein, BusU represents the actual DC bus voltage value collected, TarU represents the target output voltage value, and TarUThr represents the preset voltage deviation threshold.

[0015] Furthermore, the optimization model is:

[0016] score=Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)),

[0017] Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device.

[0018] Further, before performing action prediction on multiple rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model to obtain target recommended action information, the method further includes:

[0019] Constructing a sleep action set of the plurality of rectifier devices by using the constraint model;

[0020] The step of performing action prediction on multiple rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the dormant energy-saving model to obtain target recommended action information includes:

[0021] By using the optimization model in the sleep energy-saving model, each sleep action in the sleep action set is sequentially combined with the first characteristic parameter and the second characteristic parameter to determine a decision score for each sleep action;

[0022] The sleep action corresponding to the maximum sleep score is predicted to be target recommended action information of the plurality of rectifier devices in the next recommendation cycle.

[0023] Furthermore, the method further comprises:

[0024] If it is determined that the current powered-on number of the rectifier device is less than or equal to a preset minimum value, then the receiving of the target recommended action information is stopped.

[0025] In a second aspect, the present application provides a rectifying device processing device, comprising:

[0026] An acquisition unit, used for acquiring first characteristic parameters of the plurality of rectifying devices in a current cycle according to a preset recommended cycle during monitoring of the plurality of rectifying devices;

[0027] A training and updating unit, configured to determine a second characteristic parameter according to the first characteristic parameter; and to train and update a preset sleep energy-saving model according to the first characteristic parameter, the second characteristic parameter, and the training data of the previous cycle before the current cycle; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict target recommended action information according to multiple sleep actions in the sleep action set;

[0028] A prediction unit, configured to predict the actions of multiple rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model, and obtain target recommended action information;

[0029] The changing unit is used to change the use status of multiple rectifier devices according to the target recommended action information.

[0030] Furthermore, the training update unit includes:

[0031] A first updating module, used for determining and updating a current constraint target value and a decision score based on the first characteristic parameter and the second characteristic parameter;

[0032] The second updating module is used to train and update the constraint model in the preset hibernation energy-saving model according to the first characteristic parameter, the second characteristic parameter, the current constraint target value and the training data of the previous cycle before the current cycle; and to train and update the optimization model in the preset hibernation energy-saving model according to the first characteristic parameter, the second characteristic parameter, the decision score and the training data of the previous cycle before the current cycle.

[0033] Furthermore, the constraint model includes constraint condition information, and the constraint condition information is:

[0034] |BusU-TarU| <TarUThr,

[0035] Wherein, BusU represents the actual DC bus voltage value collected, TarU represents the target output voltage value, and TarUThr represents the preset voltage deviation threshold.

[0036] Furthermore, the optimization model is:

[0037] score=Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)),

[0038] Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device.

[0039] Furthermore, the device also includes:

[0040] A construction unit, configured to construct a sleep action set of the plurality of rectifier devices by using the constraint model before performing action prediction on the plurality of rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model to obtain target recommended action information;

[0041] The prediction unit comprises:

[0042] A first determination module is used to combine each sleep action in the sleep action set with the first characteristic parameter and the second characteristic parameter in turn through the optimization model in the sleep energy-saving model to determine a decision score for each sleep action;

[0043] The second determining module is used to predict that the sleep action corresponding to the maximum sleep score is the target recommended action information of the plurality of rectifier devices in the next recommendation period.

[0044] Furthermore, the device also includes:

[0045] A determination unit is used to stop receiving the target recommended action information if it is determined that the current number of powered-on devices of the rectifier device is less than or equal to a preset minimum value.

[0046] In a third aspect, the present application provides a server, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0047] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method described in the first aspect.

[0048] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the method described in the first aspect when executed by a processor.

[0049] The present application provides a rectifier processing method, device and server. In the process of monitoring multiple rectifiers, the first characteristic parameters of multiple rectifiers in the current cycle are obtained according to the preset recommended cycle. According to the first characteristic parameter, the second characteristic parameter is determined; and according to the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle, the preset sleep energy-saving model is trained and updated; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict the target recommended action information according to multiple sleep actions in the sleep action set. According to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model, the action of multiple rectifiers in the next recommended cycle is predicted to obtain the target recommended action information; and according to the target recommended action information, the use status of multiple rectifiers is changed. In this scheme, multiple rectifiers are monitored, and during the monitoring process, the first characteristic parameter of the rectifier is obtained, and the second characteristic parameter of the rectifier is determined according to the first characteristic parameter of the rectifier. The dormant energy-saving model is trained and updated using the first characteristic parameter and the second characteristic parameter, and the dormant energy-saving model and the first characteristic parameter and the second characteristic parameter at the current moment are used to recommend the optimal target recommended action information. Therefore, based on the actual working conditions, by dynamically adjusting the start and stop states of power modules such as rectifier equipment, the rectifier equipment can be operated in a higher efficiency range, which can improve the conversion efficiency of the power supply system more finely and intelligently, avoid the problems existing in the current mainstream solutions, achieve more refined and efficient energy saving, and solve the technical problem that the energy-saving control of the dormant energy-saving technology of the rectifier equipment is not fine enough, resulting in poor energy-saving effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0051] Figure 1 A schematic diagram of a flow chart of a rectification device processing method provided in an embodiment of the present application;

[0052] Figure 2 A schematic flow chart of another rectification device processing method provided in an embodiment of the present application;

[0053] Figure 3 A schematic diagram of the architecture of a rectification device processing method provided in an embodiment of the present application;

[0054] Figure 4 A schematic diagram of a scenario of a rectification device processing method provided in an embodiment of the present application;

[0055] Figure 5A schematic flow chart of another rectification device processing method provided in an embodiment of the present application;

[0056] Figure 6 A schematic flow chart of another rectification device processing method provided in an embodiment of the present application;

[0057] Figure 7 A structural schematic diagram of a rectifier device processing device provided in an embodiment of the present application;

[0058] Figure 8 A schematic diagram of the structure of another rectifying device processing device provided in an embodiment of the present application;

[0059] Fig. 9 A schematic diagram of the structure of a server provided in an embodiment of the present application.

[0060] The above drawings have shown clear embodiments of the present disclosure, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present disclosure in any way, but to illustrate the concepts of the present disclosure to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0061] Here, exemplary embodiments will be described in detail, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present disclosure.

[0062] At present, the rectifier equipment dormancy energy-saving technology has been widely used in sites. Its basic principle is to switch the rectifier equipment so that the rectifier equipment works in a high efficiency range, thereby achieving the purpose of system energy saving.

[0063] In one example, in the current mainstream solution, the high efficiency range of the rectifier equipment is defined by pre-set soft opening point and soft closing point. When the load rate of the rectifier equipment is higher than the soft opening point, the monitoring unit turns on one or more rectifier equipment; when the load rate of the rectifier equipment is lower than the soft closing point, the monitoring unit turns off one or more rectifier equipment. Through this control method, the load rate of the rectifier equipment is in a relatively high efficiency range. However, in actual applications, the setting of soft closing and soft opening points is performed manually, and the default soft closing point is usually 90%, and the soft opening point is 40%. The range is wide, and it is difficult to ensure customized adjustments for different rectifier equipment characteristics, which makes it difficult to implement refined energy-saving control strategies, and thus leads to poor energy-saving effects.

[0064] In one example, due to the setting of the soft closing and soft opening points, which is based on the efficiency characteristic curve of the rectifier equipment under standard working conditions, it is difficult to reflect the efficiency characteristics of the rectifier equipment under actual working conditions.

[0065] The present application provides a rectification equipment processing method, device and server, aiming to solve the above technical problems in the prior art.

[0066] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0067] Figure 1 A flow chart of a rectification device processing method provided in an embodiment of the present application is shown as follows: Figure 1 As shown, the method includes:

[0068] Step 101 : in a process of monitoring a plurality of rectifying devices, first characteristic parameters of the plurality of rectifying devices in a current cycle are obtained according to a preset recommended cycle.

[0069] Exemplarily, the executor of this embodiment may be a server. First, a plurality of different types of rectifier devices are deployed in the power supply system. During the operation of the plurality of rectifier devices, the server periodically monitors and adjusts the power supply system at certain time intervals. It should be noted that the present application uses rectifier devices as power conversion units, and other power conversion units may also be used, without limitation. In the process of monitoring the plurality of rectifier devices, the first characteristic parameters of the plurality of rectifier devices in the current cycle are obtained according to the preset recommended cycle. Among them, the first characteristic parameters include but are not limited to:

[0070] Input current InIi of each rectifier device; input voltage InUi of each rectifier device; output current OutIi of each rectifier device; output voltage OutUi of each rectifier device; module ambient temperature AmbientT; module mainboard temperature BoardT.

[0071] It should be noted that before the server obtains the first characteristic parameter, it is necessary to first perform a safety check on the operating status of the power supply system. If it is determined that the power supply system is in an abnormal state, the hibernation energy-saving control is exited. If it is determined that the power supply system is in a normal state, the first characteristic parameters of multiple rectifier devices in the current cycle are obtained.

[0072] Step 102: determine the second characteristic parameter based on the first characteristic parameter; and train and update the preset sleep energy-saving model based on the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict the target recommended action information based on multiple sleep actions in the sleep action set.

[0073] Exemplarily, the server will update the sleep energy-saving model based on the collected sample training, and use the sleep energy-saving model to make rectifier equipment start and stop recommendations for the current state of the power system, and adjust the usage status of multiple rectifier devices based on the start and stop recommendations, and then enter the next recommendation cycle.

[0074] First, according to the first characteristic parameter, the second characteristic parameter is determined, including but not limited to:

[0075] Load power LoadP = Σi (OutUi*OutIi), i represents the number of rectifier devices;

[0076] Average input voltage AvgInU = (ΣiInUi) / m, where m is the number of modules in operation;

[0077] Average output voltage AvgOutU = (ΣiOutUi) / m;

[0078] Total input current ToInI=Σi(InIi);

[0079] Total output current ToOutI = Σi (OutIi);

[0080] The number of powered-on modules of each type is ModNj;

[0081] Load change rate: LoadRatio = LoadP / LoadP of the previous cycle. Multiple cycles can be considered. If there are multiple cycles, then LoadRatio = LoadP / the sum of LoadP of multiple cycles, or LoadRatio = LoadP / the average value of LoadP of multiple cycles.

[0082] Load change value: LoadDiff = LoadP - LoadP of the previous cycle. Multiple cycles can also be considered. If there are multiple cycles, then LoadDiff = LoadP - the sum of LoadP of multiple cycles, or LoadDiff = LoadP - the average value of LoadP of multiple cycles.

[0083] Load gradient: LoadGrad, the gradient value is the slope of the least squares linear fit;

[0084] Then, according to the first characteristic parameter and the second characteristic parameter, the current constraint target value and decision score of the power system are refreshed:

[0085] The constraint target value represents: the actual busbar voltage value BusU collected;

[0086] The decision score score indicates:

[0087] score=Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)),

[0088] Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, ModOffPj represents the preset standby power of each type of rectifier device, and the standby power ModOffPj of each type of rectifier device is a user-set value, which can be determined by measurement.

[0089] And the preset sleep energy-saving model is trained and updated according to the refreshed constraint target value and decision score. Specifically, the server trains and updates the constraint model in the preset sleep energy-saving model according to the first characteristic parameter, the second characteristic parameter, the current constraint target value, and the training data of the previous cycle before the current cycle. And the optimization model in the preset sleep energy-saving model is trained and updated according to the first characteristic parameter, the second characteristic parameter, the decision score, and the training data of the previous cycle before the current cycle. Among them, the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, the constraint model is used to generate the sleep action set, and the optimization model is used to predict the target recommended action information according to multiple sleep actions in the sleep action set.

[0090] It should be noted that the constraint model uses a regression or classification algorithm to eliminate sleep change actions that may cause unreliability or insecurity. Preferably, the xgboost or svm algorithm can be used. When xgboost is used as the preferred algorithm, the parameters max_depth = 5, learning_rate = 0.2, and other algorithms and parameters can also be used. It is also recommended to introduce randomness into the constraint model to enhance the learning ability of load changes and environmental changes, that is, to invalidate the constraint model with a small probability (such as 2%).

[0091] The optimization model adopts a context-based bandit system (UCB) algorithm, and may also adopt a Gaussian process-based bandit algorithm CGP-UCB. The bandit system (UCB) algorithm is used to explore and optimize the optimal module sleep change action and improve the module conversion efficiency. Preferably, the context-based bandit (UCB) system algorithm adopts a context-dependent LinUCB algorithm or a Gaussian process-based UCB (hereinafter referred to as CGP-UCB) algorithm, wherein the LinUCB algorithm is used to adjust the utilization and exploration parameter alpha=0.5, and other algorithms and parameters may also be used.

[0092] Step 103: predict the actions of multiple rectifying devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the dormant energy-saving model to obtain target recommended action information; and change the use status of the multiple rectifying devices according to the target recommended action information.

[0093] Exemplarily, at intervals of each recommendation cycle, an optimization model is used to predict the actions of multiple rectifier devices in the next recommendation cycle based on the current first parameter information and second parameter information of the power supply system, to obtain the recommended optimal target recommended action information, and based on the target recommended action information, the usage status of the multiple rectifier devices is changed.

[0094] In an embodiment of the present application, during the monitoring process of multiple rectifier devices, the first characteristic parameters of multiple rectifier devices in the current cycle are obtained according to the preset recommended cycle. According to the first characteristic parameter, the second characteristic parameter is determined; and according to the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle, the preset sleep energy-saving model is trained and updated; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict the target recommended action information according to multiple sleep actions in the sleep action set. According to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model, the action of multiple rectifier devices in the next recommended cycle is predicted to obtain the target recommended action information; and according to the target recommended action information, the use status of multiple rectifier devices is changed. In this scheme, multiple rectifier devices are monitored, and during the monitoring process, the first characteristic parameter of the rectifier device is obtained, and the second characteristic parameter of the rectifier device is determined according to the first characteristic parameter of the rectifier device. The dormant energy-saving model is trained and updated using the first characteristic parameter and the second characteristic parameter, and the dormant energy-saving model and the first characteristic parameter and the second characteristic parameter at the current moment are used to recommend the optimal target recommended action information. Therefore, based on the actual working conditions, by dynamically adjusting the start and stop states of power modules such as rectifier equipment, the rectifier equipment can be operated in a higher efficiency range, which can improve the conversion efficiency of the power supply system more finely and intelligently, avoid the problems existing in the current mainstream solutions, achieve more refined and efficient energy saving, and solve the technical problem that the energy-saving control of the dormant energy-saving technology of the rectifier equipment is not fine enough, resulting in poor energy-saving effect.

[0095] Figure 2 A flow chart of another rectification device processing method provided in an embodiment of the present application is as follows: Figure 2 As shown, the method includes:

[0096] Step 201 : in a process of monitoring a plurality of rectifying devices, first characteristic parameters of the plurality of rectifying devices in a current cycle are obtained according to a preset recommended cycle.

[0097] For example, this step can be referred to Figure 1 Step 101 in the above is not described in detail.

[0098] Step 202: Determine and update the current constraint target value and decision score based on the first characteristic parameter and the second characteristic parameter.

[0099] For example, this step can be referred to Figure 1 Step 102 in the above is not described in detail.

[0100] Step 203: train and update the constraint model in the preset sleep energy-saving model based on the first characteristic parameter, the second characteristic parameter, the current constraint target value, and the training data of the previous cycle before the current cycle; and train and update the optimization model in the preset sleep energy-saving model based on the first characteristic parameter, the second characteristic parameter, the decision score, and the training data of the previous cycle before the current cycle.

[0101] Among them, the training data includes historical first characteristic parameters, second characteristic parameters, and historical recommended action information. The preset sleep energy-saving model includes a constraint model and an optimization model. The constraint model is used to generate a sleep action set, and the optimization model is used to predict target recommended action information based on multiple sleep actions in the sleep action set.

[0102] In one example, the constraint model includes constraint condition information, and the constraint condition information is:

[0103] |BusU-TarU| <TarUThr;

[0104] Among them, BusU represents the actual DC bus voltage value collected, TarU represents the target output voltage value, and TarUThr represents the preset voltage deviation threshold.

[0105] In an example, the optimization model is:

[0106] score=Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj))

[0107] Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device.

[0108] Exemplarily, the working principle of the server training and updating dormant energy-saving model is described in detail as follows: the server maintains an optimization sample set Memory1 with a length of L1, and a constraint sample set Memory2 with a length of L2. These sample sets are updated in real time with the environment and are used to train and update the optimization model and the constraint model. At each recommendation cycle, we add new samples to Memory1 and Memory2. If the length of Memory1 is greater than L1 after adding new samples, the oldest sample in Memory1 is removed. If the length of Memory2 is greater than L2 after adding new samples, the oldest sample in Memory2 is removed. Preferably, L1=L2=5000.

[0109] Therefore, at each recommendation cycle, the optimization sample set and constraint sample set are updated according to the current state of the power system (i.e., the first characteristic parameter and the second characteristic parameter), and the optimization model is updated by training with the new optimization sample set, and the constraint model is updated by training with the new constraint sample set. Specifically:

[0110] a) Inputting the current first characteristic parameter and the second characteristic parameter of the power supply system.

[0111] b) According to the current first characteristic parameter and second characteristic parameter, calculate the decision score score, that is, the conversion efficiency score, if score = Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)), where score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device. Obviously, the currently calculated decision score is an evaluation of the result of the last decision.

[0112] c) Using part of the first characteristic parameter and the second characteristic parameter of the power supply system in the last recommended cycle and the sleep action of the last decision as characteristic values, and the currently calculated decision score as the target value, as new samples to update the optimized sample set; using part of the first characteristic parameter and the second characteristic parameter of the power supply system in the last recommended cycle and the sleep action of the last decision as characteristic values, and the current average output voltage as the target output voltage value, as new samples to update the constrained sample set.

[0113] d) using the updated optimization sample set to train and update the optimization model; using the updated constraint sample set to train and update the constraint model; wherein the constraint model includes constraint condition information, and the constraint condition information is:

[0114] |BusU-TarU| <TarUThr;

[0115] BusU represents the actual DC bus voltage value collected, TarU represents the target output voltage value, and TarUThr represents the preset voltage deviation threshold. When it is predicted that the power-on, power-off or unchanged action will cause |BusU-TarU| to be greater than the voltage deviation threshold, this sleep action is prohibited.

[0116] Optimization model score = Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)),

[0117] Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device.

[0118] Step 204: construct a sleep action set for multiple rectifier devices through a constraint model.

[0119] It should be noted that 1) the constraint model can also be simplified and replaced by rules. For example, when the shutdown action changes and the rated current of the remaining rectifier equipment is less than the running current of the rectifier equipment before shutdown, the rule prohibits shutdown. At this time, the entire power system still performs well under fixed load.

[0120] 2) The prediction type of the constraint model may also be changed, for example, from predicted voltage to predicted load rate. Specifically, when the predicted shutdown action changes and causes the predicted load rate to be greater than or equal to 100%, shutdown is prohibited.

[0121] 3) You can also remove the constraint model and incorporate the constraint information into the optimization model through weighting. Specifically, when the voltage or load rate meets the constraint information, the decision score is still:

[0122] Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)).

[0123] When the voltage or load rate does not meet the constraint information, the decision score is:

[0124] Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj))-99,

[0125] Or, Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj))–λ*|BusU-TarU|. This can reduce the score of information that does not meet the constraints and avoid selecting bad target recommended action information.

[0126] Exemplarily, the server constructs a sleep action set of the rectifier device through a constraint model. In the sleep action set, one rectifier device can be turned on or off at a time, or multiple rectifier devices can be turned on or off at a time. Preferably, the on and off state of the rectifier device can be used as an action, and the on and off states of all rectifier devices are arranged and combined to obtain a sleep action set, which includes: {ModN1+1, ModN1-1, ... ModNn+1, ModNn-1, 0}, wherein ModN1, ..., ModNn represent rectifier devices of different power categories, such as rectifier devices of high-efficiency power, rectifier devices of ordinary power, etc., n represents the number of categories of all rectifier devices, and 0 represents that all categories of rectifier devices do not change from start to stop. Among the rectifier devices of all power categories, the control effect is gentler if only one rectifier device of one power category is turned on or off at a time, that is, only one rectifier device of one power category is turned on or off, or the rectifier device remains unchanged, that is, neither turned on or off. The above is the preferred suggestion; it is also possible to turn on two rectifier devices of one power category, or turn off two rectifier devices, etc.; or, it is also possible to turn on and off two rectifier devices of two power categories at the same time, and there is no limitation on this.

[0127] For example, there are two rectifier devices, a and b, both of which are rectifier devices of normal power. The sleep action set includes: {a is on, b is off; a is off, b is on; a is on, b is on; a is off, b is off; 0}. If a is on and b is off, then the a off b on action changes the rectifier devices of two power categories at once.

[0128] Alternatively, in the sleep action set of the rectifier device, the number of rectifier devices turned on may be used as an action. For example, the action set of the rectifier coefficients of a device with only 8 high-efficiency rectifier devices may be {0, 1, 2, ..., 8}, and there is no limitation on this.

[0129] Step 205 : by using the optimization model in the sleep energy-saving model, each sleep action in the sleep action set is sequentially combined with the first characteristic parameter and the second characteristic parameter to determine a decision score for each sleep action.

[0130] Exemplarily, the server traverses the action set in the sleep action set, combines part of the first and second characteristic parameters of the above power system with the sleep action as input, and uses the optimization model in the sleep energy-saving model to predict the decision score of each sleep action.

[0131] Step 206: Predict the sleep action corresponding to the maximum sleep score as target recommended action information for multiple rectifier devices in the next recommendation period.

[0132] For example, the server may predict, based on the decision score, that the dormant action corresponding to the maximum decision score is the predicted and recommended optimal action.

[0133] Step 207: Change the usage status of multiple rectifier devices according to the target recommended action information.

[0134] Exemplarily, the server changes the usage status of multiple rectifying devices according to the target recommended action information. For example, the number of rectifying devices is 3, namely a, b and c, and the target recommended action information is a on, b off and c on, then the server changes the usage status of multiple rectifying devices in the next cycle according to the target recommended action information "a on, b off and c on".

[0135] Step 208: If it is determined that the current number of rectifier devices turned on is less than or equal to a preset minimum value, stop receiving target recommended action information.

[0136] Exemplarily, if the server determines that the current number of powered-on rectifier devices is ≤ the preset minimum value ModNmin, the server will no longer accept the recommended shutdown action; if the server determines that the current number of powered-on devices has reached the maximum, the server will no longer accept the recommended power-on action.

[0137] In the embodiment of the present application, in the process of monitoring multiple rectifier devices, the first characteristic parameters of multiple rectifier devices in the current cycle are obtained according to the preset recommended cycle. Based on the first characteristic parameter and the second characteristic parameter, the current constraint target value and the decision score are determined and updated. According to the first characteristic parameter, the second characteristic parameter, the current constraint target value and the training data of the previous cycle before the current cycle, the constraint model in the preset dormant energy-saving model is trained and updated; and according to the first characteristic parameter, the second characteristic parameter, the decision score and the training data of the previous cycle before the current cycle, the optimization model in the preset dormant energy-saving model is trained and updated. The dormant action set of multiple rectifier devices is constructed by the constraint model. Therefore, based on the actual working conditions, by dynamically adjusting the start and stop states of power modules such as rectifier devices, the rectifier device can be operated in a higher efficiency range, which can improve the conversion efficiency of the power supply system more finely and intelligently, avoid the problems existing in the current mainstream solution, achieve more refined and efficient energy saving, and solve the technical problem that the energy-saving control of the dormant energy-saving technology of the rectifier device is not fine enough, resulting in poor energy-saving effect. Through the optimization model in the dormant energy-saving model, each dormant action in the dormant action set is combined with the first characteristic parameter and the second characteristic parameter in turn to determine the decision score of each dormant action. The sleep action corresponding to the predicted maximum sleep score is the target recommended action information of multiple rectifier devices in the next recommendation cycle. According to the target recommended action information, the use status of multiple rectifier devices is changed. If it is determined that the current number of rectifier devices turned on is less than or equal to the preset minimum value, the target recommended action information is stopped.

[0138] For example, Figure 3 A schematic diagram of the architecture of a rectifier device processing method provided in an embodiment of the present application, such as Figure 3 As shown in the figure, the architecture includes: server, rectifier, battery and load. The rectifier realizes the conversion from AC to DC and provides power for the DC load. The battery is connected to the DC busbar in parallel with the rectifier as an energy storage component. The rectifier charges the battery at ordinary times. When the city power is cut off and the rectifier has no output, the battery takes on the load output. The server monitors and manages the operation of the entire power system and provides a human-machine interface.

[0139] Figure 4 A schematic diagram of a rectifier device processing method provided in an embodiment of the present application, such as Figure 4 As shown, the rectifier is the power conversion component in the power system and is also the core component that affects the efficiency of the entire power system. Under the rated input voltage, Figure 4 The typical efficiency curve of the rectifier is shown. From the efficiency curve, we can see that the conversion efficiency of the rectifier is not fixed and will be affected by specific working conditions such as load and input voltage.

[0140] Figure 5 A flow chart of another rectification device processing method provided in an embodiment of the present application is as follows: Figure 5 As shown, the method includes: monitoring the power supply system (i.e., the power supply system is a dormant energy-saving system) through a server, a recommendation / optimization system is deployed in the server, the recommendation / optimization system includes a training subsystem and a decision subsystem, and the setting parameters include a preset voltage deviation threshold, a target output voltage value, a preset minimum value, etc., and the state parameters include first parameter information and second parameter information, etc. The training subsystem trains the dormant energy-saving model based on samples consisting of environmental information (i.e., first parameter information and second parameter information) of the power supply system, actions, and decision scores. The decision subsystem predicts and recommends the optimal target recommended action information based on the dormant energy-saving model and the environmental information of the power supply system.

[0141] Figure 6 A flow chart of another rectification device processing method provided in an embodiment of the present application is as follows: Figure 6 As shown, the method includes: inputting parameters, where the parameters refer to the first characteristic parameters; safety refers to performing a safety check on the operating status of the power supply system. If it is determined that the power supply system is in an abnormal state, then exiting the sleep energy-saving control; if it is determined that the power supply system is in a normal state, then performing feature processing on the input parameters, and the feature processing includes calculating the second characteristic parameter, refreshing the current constraint target value and decision score of the power supply system, etc.; based on the xgboost model, obtaining a sleep action set that meets the constraint condition information; if the optional actions in the sleep action set are less than or equal to 1, then directly outputting a unique action; if the optional actions are greater than 1, then outputting the recommended action through the LinUCB model, that is, the target recommended action information.

[0142] Figure 7 A structural diagram of a rectifier device processing device provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the device comprises:

[0143] The acquisition unit 31 is used to acquire the first characteristic parameters of the plurality of rectifying devices in the current cycle according to a preset recommended cycle during the monitoring process of the plurality of rectifying devices.

[0144] The training update unit 32 is used to determine the second characteristic parameter based on the first characteristic parameter; and to train and update the preset sleep energy-saving model based on the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict the target recommended action information based on multiple sleep actions in the sleep action set.

[0145] The prediction unit 33 is used to predict the actions of multiple rectifier devices in the next recommendation period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model to obtain target recommended action information.

[0146] The changing unit 34 is used to change the use status of the plurality of rectifier devices according to the target recommended action information.

[0147] The device of this embodiment can execute the technical solution in the above method. Its specific implementation process and technical principles are the same and will not be repeated here.

[0148] Figure 8 A schematic diagram of the structure of another rectifier device processing device provided in an embodiment of the present application, Figure 7 Based on the embodiment shown, Figure 8 As shown, the training update unit 32 includes:

[0149] The first updating module 321 is used to determine and update the current constraint target value and decision score based on the first characteristic parameter and the second characteristic parameter.

[0150] The second update module 322 is used to train and update the constraint model in the preset sleep energy-saving model according to the first characteristic parameter, the second characteristic parameter, the current constraint target value and the training data of the previous cycle before the current cycle; and to train and update the optimization model in the preset sleep energy-saving model according to the first characteristic parameter, the second characteristic parameter, the decision score and the training data of the previous cycle before the current cycle.

[0151] In one example, the constraint model includes constraint condition information, and the constraint condition information is:

[0152] |BusU-TarU| <TarUThr,

[0153] Among them, BusU represents the actual DC bus voltage value collected, TarU represents the target output voltage value, and TarUThr represents the preset voltage deviation threshold.

[0154] In an example, the optimization model is:

[0155] score=Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)),

[0156] Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device.

[0157] In one example, the device further includes:

[0158] The construction unit 41 is used to construct a sleep action set of multiple rectifier devices through a constraint model before predicting the actions of multiple rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model to obtain target recommended action information.

[0159] The prediction unit 33 includes:

[0160] The first determination module 331 is used to combine each sleep action in the sleep action set with the first characteristic parameter and the second characteristic parameter in turn through the optimization model in the sleep energy-saving model to determine the decision score of each sleep action.

[0161] The second determination module 332 is used to predict that the sleep action corresponding to the maximum sleep score is the target recommended action information of multiple rectifier devices in the next recommendation period.

[0162] In one example, the device further includes:

[0163] The determination unit 42 is configured to stop receiving target recommended action information if it is determined that the current number of rectifier devices turned on is less than or equal to a preset minimum value.

[0164] The device of this embodiment can execute the technical solution in the above method. Its specific implementation process and technical principles are the same and will not be repeated here.

[0165] Fig. 9A schematic diagram of the structure of a server provided in an embodiment of the present application, such as Fig. 9 As shown, the server includes: a memory 51 and a processor 52.

[0166] The memory 51 stores a computer program that can be executed on the processor 52 .

[0167] The processor 52 is configured to execute the method provided in the above embodiment.

[0168] The server further includes a receiver 53 and a transmitter 54. The receiver 53 is used to receive instructions and data sent by an external device, and the transmitter 54 is used to send instructions and data to the external device.

[0169] The embodiment of the present application also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the server, the server can execute the method provided by the above embodiment.

[0170] An embodiment of the present application also provides a computer program product, which includes: a computer program, which is stored in a readable storage medium, and at least one processor of the server can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the server executes the solution provided by any of the above embodiments.

[0171] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0172] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A rectification equipment processing method, characterized in that: include: In the process of monitoring the plurality of rectifying devices, obtaining first characteristic parameters of the plurality of rectifying devices in a current cycle according to a preset recommended cycle; According to the first characteristic parameter, a second characteristic parameter is determined; and according to the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle, a preset sleep energy-saving model is trained and updated; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict the target recommended action information according to multiple sleep actions in the sleep action set; According to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model, the action of multiple rectifying devices in the next recommended cycle is predicted to obtain target recommended action information; and according to the target recommended action information, the use status of multiple rectifying devices is changed.

2. The method according to claim 1, characterized in that The training and updating of the preset dormancy energy-saving model according to the first characteristic parameter, the second characteristic parameter and the training data of the previous cycle before the current cycle includes: Based on the first characteristic parameter and the second characteristic parameter, determine and update the current constraint target value and decision score; According to the first characteristic parameter, the second characteristic parameter, the current constraint target value and the training data of the previous cycle before the current cycle, the constraint model in the preset hibernation energy-saving model is trained and updated; and according to the first characteristic parameter, the second characteristic parameter, the decision score and the training data of the previous cycle before the current cycle, the optimization model in the preset hibernation energy-saving model is trained and updated.

3. The method according to claim 2, characterized in that The constraint model includes constraint condition information, and the constraint condition information is: |BusU-TarU| <TarUThr, Wherein, BusU represents the actual DC bus voltage value collected, TarU represents the target output voltage value, and TarUThr represents the preset voltage deviation threshold.

4. The method according to claim 2, characterized in that The optimization model is: score=Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)), Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device.

5. The method according to claim 1, characterized in that: Before performing action prediction on multiple rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model to obtain target recommended action information, the method further includes: Constructing a sleep action set of the plurality of rectifier devices by using the constraint model; The step of performing action prediction on multiple rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the dormant energy-saving model to obtain target recommended action information includes: By using the optimization model in the sleep energy-saving model, each sleep action in the sleep action set is sequentially combined with the first characteristic parameter and the second characteristic parameter to determine a decision score for each sleep action; The sleep action corresponding to the maximum sleep score is predicted to be target recommended action information of the plurality of rectifier devices in the next recommendation cycle.

6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: If it is determined that the current powered-on number of the rectifier device is less than or equal to a preset minimum value, then the receiving of the target recommended action information is stopped.

7. A rectifying equipment processing device, characterized in that: include: An acquisition unit, used for acquiring first characteristic parameters of the plurality of rectifying devices in a current cycle according to a preset recommended cycle during monitoring of the plurality of rectifying devices; A training and updating unit, configured to determine a second characteristic parameter according to the first characteristic parameter; and to train and update a preset sleep energy-saving model according to the first characteristic parameter, the second characteristic parameter, and the training data of the previous cycle before the current cycle; wherein the training data includes the historical first characteristic parameter, the second characteristic parameter, and the historical recommended action information, and the preset sleep energy-saving model includes a constraint model and an optimization model, the constraint model is used to generate a sleep action set, and the optimization model is used to predict target recommended action information according to multiple sleep actions in the sleep action set; A prediction unit, configured to predict the actions of multiple rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model, and obtain target recommended action information; The changing unit is used to change the use status of multiple rectifier devices according to the target recommended action information.

8. The device according to claim 7, characterized in that The training update unit comprises: A first updating module, used for determining and updating a current constraint target value and a decision score based on the first characteristic parameter and the second characteristic parameter; The second updating module is used to train and update the constraint model in the preset hibernation energy-saving model according to the first characteristic parameter, the second characteristic parameter, the current constraint target value and the training data of the previous cycle before the current cycle; and to train and update the optimization model in the preset hibernation energy-saving model according to the first characteristic parameter, the second characteristic parameter, the decision score and the training data of the previous cycle before the current cycle.

9. The device according to claim 8, characterized in that The constraint model includes constraint condition information, and the constraint condition information is: |BusU-TarU| <TarUThr, Wherein, BusU represents the actual DC bus voltage value collected, TarU represents the target output voltage value, and TarUThr represents the preset voltage deviation threshold.

10. The device according to claim 8, characterized in that The optimization model is: score=Σi(OutUi*OutIi) / (Σi(InUi*InIi)+Σj(kj*ModOffPj)), Among them, score represents the decision score output by the optimization model, OutUi represents the output voltage of each rectifier device, OutIi represents the output current of each rectifier device, InUi represents the input voltage of each rectifier device, InIi represents the input current of each rectifier device, Kj represents the number of standby rectifier devices of the jth type collected, and ModOffPj represents the preset standby power of each type of rectifier device.

11. The device according to claim 7, characterized in that The device also includes: A construction unit, configured to construct a sleep action set of the plurality of rectifier devices by using the constraint model before performing action prediction on the plurality of rectifier devices in the next recommended period according to the first characteristic parameter, the second characteristic parameter and the sleep energy-saving model to obtain target recommended action information; The prediction unit comprises: A first determination module is used to combine each sleep action in the sleep action set with the first characteristic parameter and the second characteristic parameter in turn through the optimization model in the sleep energy-saving model to determine a decision score for each sleep action; The second determining module is used to predict that the sleep action corresponding to the maximum sleep score is the target recommended action information of the plurality of rectifier devices in the next recommendation period.

12. The device according to any one of claims 7 to 11, characterized in that: The device also includes: A determination unit is used to stop receiving the target recommended action information if it is determined that the current number of powered-on devices of the rectifier device is less than or equal to a preset minimum value.

13. A server, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

15. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images