Intelligent collaborative power consumption regulation and control method, apparatus and system for source-grid-load-storage, electronic device and storage medium

Abstract

The present disclosure relates to the technical field of intelligent monitoring and management of power systems, and specifically relates to an intelligent collaborative power consumption regulation and control method, apparatus and system for source-grid-load-storage, an electronic device and a storage medium. Said system comprises an energy regulation and control center and energy regulation and control units provided in microgrids; the energy regulation and control units use a temporal attention mechanism-based LRCN dual-layer network combined model to predict power consumption amounts, so as to generate power consumption surpluses and shortages within a future preset time; and on the basis of the power consumption surpluses and shortages and latest current electricity prices of the microgrids, the energy regulation and control center uses a fusion multi-objective algorithm based on a Pareto front curve and a fuzzy algorithm to generate a microgrid collaborative power consumption regulation and control solution, and sends the regulation and control solution to the energy regulation and control units for execution, so as to ensure the balance of energy supply and demand of the microgrids. Therefore, the present disclosure achieves efficient, intelligent and refined energy management for microgrid clusters, reducing energy consumption and costs, and providing solid support for sustainable development of microgrids.

Description

Source-grid-load-storage collaborative power consumption intelligent regulation method, device, system, electronic equipment and storage medium TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of intelligent monitoring and management of power systems, in particular to a source-grid-load-storage collaborative power consumption intelligent regulation method, device, system, electronic equipment and storage medium. BACKGROUND

[0002] With the adjustment of global energy structure and the rapid development of renewable energy, microgrids, as a new, flexible and efficient energy supply system and an important part of distributed energy systems, play an increasingly important role in energy supply. A microgrid cluster is composed of multiple microgrids, which can realize energy complementation and sharing, improve energy utilization efficiency, and reduce energy consumption and cost. However, in actual operation, microgrid clusters face challenges in energy supply and demand balance and efficient utilization. The imbalance of energy supply and demand within the microgrid cluster and low energy utilization efficiency have become bottlenecks restricting the further development of microgrids.

[0003] Firstly, the energy supply and demand in the microgrid cluster is affected by various factors, including weather changes, load fluctuations, energy prices, etc. Traditional microgrid energy management mainly relies on manual scheduling and fixed rule-based automated control, which cannot quickly respond to changes in energy supply and demand. This approach is difficult to adapt to rapidly changing energy supply and demand and complex external environmental factors. In particular, in microgrids with a high proportion of renewable energy, the intermittency and instability of renewable energy make it more difficult to balance energy supply and demand.

[0004] Secondly, the lack of energy interaction and coordination between microgrids in the microgrid cluster is also an important reason for low energy utilization efficiency. Each microgrid often only focuses on its own energy supply and demand, ignoring cooperation with other microgrids. This leads to waste of energy resources and reduction of energy utilization efficiency, and also increases the operating cost and risk of the microgrid cluster.

[0005] How to achieve efficient, intelligent and fine-grained management of energy in the microgrid cluster, improve energy utilization efficiency, reduce energy consumption and cost, and provide strong support for the sustainable development of microgrids is a problem that needs to be solved. SUMMARY

[0006] To solve the problems in the related art, the embodiments of the present disclosure provide a source-grid-load-storage collaborative power consumption intelligent regulation method, device, system, electronic equipment and storage medium.

[0007] In a first aspect, the disclosure provides a source-network-load-storage collaborative intelligent power regulation system, which comprises an energy regulation center and an energy regulation unit arranged in each micro-grid in a micro-grid cluster. The energy regulation center is in communication connection with the energy regulation units in each micro-grid. The energy regulation units in each micro-grid in the micro-grid cluster are in communication connection.

[0008] The energy regulation unit in any micro-grid in the micro-grid cluster is configured to:

[0009] obtain historical power consumption information and meteorological parameters in a future preset time of the any micro-grid; the historical power consumption information comprises historical electrical load and historical thermal load, and the meteorological parameters comprise one or more of the following parameters: illumination intensity, temperature, wind speed, wind direction, and relative humidity;

[0010] jointly predict the power consumption of the any micro-grid in the future preset time according to the historical power consumption information and the meteorological parameters using a LRCN double-layer network combination model based on a time attention mechanism; the LRCN double-layer network combination model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module, and a multi-source data feature sharing layer module; the jointly predicting the power consumption of the any micro-grid in the future preset time according to the historical power consumption information and the meteorological parameters using the LRCN double-layer network combination model based on the time attention mechanism comprises: processing the historical electrical load using a first load feature extraction channel to obtain a feature representation of the historical electrical load; processing the historical thermal load using a second load feature extraction channel to obtain a feature representation of the historical thermal load; processing the temperature in the meteorological parameters, the feature representation of the historical electrical load, and the feature representation of the historical thermal load using the multi-source data feature sharing layer module to obtain an electrical load prediction value and a thermal load prediction value of the any micro-grid in the future preset time; and obtaining the power consumption of the any micro-grid in the future preset time according to the electrical load prediction value and the thermal load prediction value;

[0011] determining the power generation of the any micro-grid in the future preset time according to the meteorological parameters; determining the power surplus / deficit of the any micro-grid in the future preset time according to the current energy storage capacity in the any micro-grid, the power generation in the future preset time, and the power consumption in the future preset time; and sending the power surplus / deficit in the future preset time to the energy regulation center;

[0012] obtaining the latest current price of the any micro-grid according to the power generation cost, the transmission cost, and the active power loss of the any micro-grid; and sending the latest current price to the energy regulation center.

[0013] The energy regulation center is configured to generate a micro-grid cooperative power consumption regulation scheme using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power consumption surplus and deficiency of each micro-grid in the micro-grid cluster within the future preset time and the latest current electricity price, and send the micro-grid cooperative power consumption regulation scheme to the energy regulation unit in any micro-grid.

[0014] The energy regulation unit is further configured to regulate power consumption according to the micro-grid cooperative power consumption regulation scheme.

[0015] According to an embodiment of the present disclosure, the any micro-grid further comprises a power consumption data management system DMS configured to store, manage and maintain the historical electrical load and the historical thermal load, and the historical power consumption information of the any micro-grid is obtained, comprising:

[0016] The historical power consumption information is obtained through interface communication with the power consumption data management system DMS.

[0017] According to an embodiment of the present disclosure, the first load feature extraction channel comprises a first convolutional neural network CNN module, a first long short-term memory network LSTM module and a first efficient multi-scale attention module, and the historical electrical load is processed using the first load feature extraction channel to obtain a feature representation of the historical electrical load, comprising:

[0018] The local features of the historical electrical load are extracted using the first convolutional neural network CNN module through convolution operation to obtain a feature map of the historical electrical load.

[0019] The feature map of the historical electrical load is processed using the first long short-term memory network LSTM module to capture long-term and short-term dependencies in the historical electrical load based on the characteristics of the recurrent connection to obtain a hidden state vector of the historical electrical load.

[0020] The hidden state vector of the historical electrical load is processed using the first efficient multi-scale attention module to obtain a feature representation of the historical electrical load after attention mechanism weighting by assigning different weights to different features.

[0021] The second load feature extraction channel comprises a second convolutional neural network CNN module, a second long short-term memory network LSTM module and a second efficient multi-scale attention module, and the historical thermal load is processed using the second load feature extraction channel to obtain a feature representation of the historical thermal load, comprising:

[0022] extracting local features of the historical thermal load by using the second convolutional neural network (CNN) module through a convolution operation to obtain a feature map of the historical thermal load;

[0023] processing the feature map of the historical thermal load by using the second long short-term memory (LSTM) module to capture long-term and short-term dependencies in the historical thermal load based on the characteristics of a recurrent connection to obtain a hidden state vector of the historical thermal load;

[0024] processing the hidden state vector of the historical thermal load by using the second efficient multi-scale attention module to obtain a feature representation of the historical thermal load after attention mechanism weighting by giving different weights to different features.

[0025] According to an embodiment of the present disclosure, the energy regulation unit is further configured to:

[0026] obtain a current actual electrical load and a current actual thermal load;

[0027] obtain an electrical load deviation according to the current actual electrical load and the electrical load prediction value;

[0028] obtain a thermal load deviation according to the current actual thermal load and the thermal load prediction value;

[0029] when the electrical load deviation and / or the thermal load deviation do not meet preset requirements, performing parameter migration on the LRCN double-layer network combination model based on a time attention mechanism by using an online learning framework based on transfer learning, including:

[0030] if a difference between the average value of the historical electrical load and the current actual electrical load is greater than a preset electrical load difference threshold, updating first load feature extraction channel parameters;

[0031] if a difference between the average value of the historical thermal load and the current actual thermal load is greater than a preset thermal load difference threshold, updating second load feature extraction channel parameters;

[0032] obtaining a current temperature and a historical temperature, and if a difference between the average value of the historical temperature and the current temperature is greater than a preset temperature difference threshold, updating the first load feature extraction channel parameters and the second load feature extraction channel parameters.

[0033] According to an embodiment of the present disclosure, the meteorological parameters further include: horizontal surface solar radiation per unit time;

[0034] the power generation in the future preset time is calculated by the following formula: E = H x P x K x T;

[0035] Wherein, E is the power generation in the future preset time, H is the horizontal surface solar radiation in the unit time, P is the system installation capacity, K is the comprehensive efficiency coefficient, and T is the future preset time.

[0036] According to an embodiment of the present disclosure, the determination of the power surplus or deficit of the any micro-grid in the future preset time according to the current energy storage capacity in the any micro-grid and the power generation in the future preset time and the power consumption in the future preset time comprises:

[0037] The power surplus or deficit in the future preset time is calculated by the following formula: Q=A+B-C.

[0038] Wherein, Q is the power surplus or deficit in the future preset time, A is the current energy storage capacity in the any micro-grid, B is the power generation in the future preset time, and C is the power consumption in the future preset time.

[0039] When Q is positive, the power surplus or deficit in the future preset time indicates that the any micro-grid is in a power surplus state; when Q is negative, the power surplus or deficit in the future preset time indicates that the any micro-grid is in a power deficit state; and when Q is zero, the power surplus or deficit in the future preset time indicates that the any micro-grid is in a power balance state.

[0040] According to an embodiment of the present disclosure, the acquisition of the latest current price of the any micro-grid according to the power generation cost, the transmission cost and the active power loss of the any micro-grid comprises:

[0041] The power generation cost of the any micro-grid is calculated based on the active power of the any micro-grid, the current price, the total discharge power of the any micro-grid and the demand side response power consumption of the any micro-grid in a specified time.

[0042] The transmission cost is determined based on the active power from the any micro-grid to the transmission line in the specified time.

[0043] The active power loss in the specified time is determined based on the total transmission line topology level, the transmission line resistance and the transmission line current.

[0044] The latest current price is the price that satisfies the conditions that the power generation cost is less than a first preset constant, the active power loss is less than a second preset constant and the transmission cost is the minimum.

[0045] According to an embodiment of the present disclosure, the generation of the micro-grid cooperative power consumption regulation scheme using the fusion multi-objective algorithm based on the Pareto frontier curve and the fuzzy algorithm according to the power surplus or deficit in the future preset time and the latest current price of each micro-grid in the micro-grid cluster comprises:

[0046] The fuzzy algorithm is used to fuzzify the electricity surplus / deficit and the latest current electricity price; the fuzzified electricity surplus / deficit and the latest current electricity price are converted into membership values ​​in a fuzzy set using a predefined membership function;

[0047] Based on preset parameters, corresponding weights are assigned to multiple predefined optimization objectives; the optimization objectives include: minimizing electricity costs, maximizing energy utilization efficiency, or reducing environmental impact.

[0048] Based on the multiple optimization objectives and the weights corresponding to each optimization objective, a multi-objective optimization algorithm is used to iteratively calculate the membership values ​​in the fuzzy set to generate multiple preliminary power control schemes. The multi-objective optimization algorithm includes: particle swarm optimization algorithm or genetic algorithm.

[0049] Evaluate multiple objective function values ​​for each alternative power control scheme and mark the position of each alternative power control scheme on the Pareto front curve;

[0050] Combining the Pareto front curve and the membership values ​​in the fuzzy set, the microgrid collaborative power control scheme is selected or generated through a preset fusion strategy; the preset fusion strategy includes weighted fusion.

[0051] Secondly, this disclosure provides a source-grid-load-storage coordinated intelligent power control method, applied to an energy control unit. The energy control unit is located within each microgrid in a microgrid cluster and is communicatively connected to the energy control center within the microgrid cluster. The energy control units within each microgrid in the microgrid cluster are communicatively connected to each other. The intelligent control method includes:

[0052] The energy control unit in any microgrid within the microgrid cluster acquires historical electricity consumption information and meteorological parameters for a future preset time period for any microgrid; the historical electricity consumption information includes historical electrical load and historical heat load, and the meteorological parameters include one or more of the following parameters: light intensity, temperature, wind speed, wind direction, and relative humidity;

[0053] According to the historical power consumption information and the meteorological parameters, a LRCN double-layer network combined model based on a time attention mechanism is used to jointly predict power consumption of the any micro-grid in the future preset time; the LRCN double-layer network combined model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module and a multi-source data feature sharing layer module; according to the historical power consumption information and the meteorological parameters, the LRCN double-layer network combined model based on the time attention mechanism is used to jointly predict power consumption of the any micro-grid in the future preset time, which comprises: using a first load feature extraction channel to process the historical electric load to obtain a feature representation of the historical electric load; using a second load feature extraction channel to process the historical thermal load to obtain a feature representation of the historical thermal load; using the multi-source data feature sharing layer module to process temperature in the meteorological parameters, the feature representation of the historical electric load and the feature representation of the historical thermal load to obtain an electric load prediction value and a thermal load prediction value of the any micro-grid in the future preset time; and obtaining power consumption of the any micro-grid in the future preset time according to the electric load prediction value and the thermal load prediction value.

[0054] According to the meteorological parameters, power generation of the any micro-grid in the future preset time is determined; according to the current energy storage capacity in the any micro-grid, power generation in the future preset time and power consumption in the future preset time, power surplus or deficiency of the any micro-grid in the future preset time is determined; and the power surplus or deficiency of the any micro-grid in the future preset time is sent to the energy regulation center.

[0055] According to the power generation cost, transmission cost and active power loss of the any micro-grid, a latest current price of the any micro-grid is obtained; and the latest current price is sent to the energy regulation center.

[0056] A micro-grid cooperative power consumption regulation scheme returned by the energy regulation center is received.

[0057] According to the micro-grid cooperative power consumption regulation scheme, power consumption regulation is performed.

[0058] According to an embodiment of the present disclosure, the first load feature extraction channel comprises a first convolutional neural network (CNN) module, a first long short-term memory (LSTM) module and a first efficient multi-scale attention module, and the first load feature extraction channel is used to process the historical electric load to obtain a feature representation of the historical electric load, which comprises:

[0059] The first CNN module is used to extract local features of the historical electric load through convolution operation to obtain a feature map of the historical electric load.

[0060] processing the feature map of the historical electric load using the first long short-term memory network (LSTM) module, capturing long-term and short-term dependencies in the historical electric load through the characteristics based on the loop connection, obtaining a hidden state vector of the historical electric load;

[0061] processing the hidden state vector of the historical electric load using the first high-efficiency multi-scale attention module, obtaining a feature representation of the historical electric load after being weighted by the attention mechanism by assigning different weights to different features;

[0062] The second load feature extraction channel includes a second convolutional neural network (CNN) module, a second long short-term memory network (LSTM) module, and a second high-efficiency multi-scale attention module, and the historical thermal load is processed using the second load feature extraction channel to obtain a feature representation of the historical thermal load, including:

[0063] extracting local features of the historical thermal load through convolution operation using the second convolutional neural network (CNN) module, obtaining a feature map of the historical thermal load;

[0064] processing the feature map of the historical thermal load using the second long short-term memory network (LSTM) module, capturing long-term and short-term dependencies in the historical thermal load through the characteristics based on the loop connection, obtaining a hidden state vector of the historical thermal load;

[0065] processing the hidden state vector of the historical thermal load using the second high-efficiency multi-scale attention module, obtaining a feature representation of the historical thermal load after being weighted by the attention mechanism by assigning different weights to different features.

[0066] According to an embodiment of the present disclosure, the intelligent regulation method further includes:

[0067] obtaining a current actual electric load and a current actual thermal load;

[0068] obtaining an electric load deviation according to the current actual electric load and the electric load prediction value;

[0069] obtaining a thermal load deviation according to the current actual thermal load and the thermal load prediction value;

[0070] when the electric load deviation and / or the thermal load deviation do not meet the preset requirements, performing parameter migration on the LRCN double-layer network combination model based on the time attention mechanism using an online learning framework based on transfer learning, including:

[0071] if the difference between the average value of the historical electric load and the current actual electric load is greater than a preset electric load difference threshold, updating the first load feature extraction channel parameters;

[0072] If a difference between the average value of the historical thermal load and the current actual thermal load is greater than a preset thermal load difference threshold, the second load feature extraction channel parameter is updated;

[0073] The current temperature and the historical temperature are obtained, and if a difference between the average value of the historical temperature and the current temperature is greater than a preset temperature difference threshold, the first load feature extraction channel parameter and the second load feature extraction channel parameter are updated.

[0074] According to an embodiment of the present disclosure, the meteorological parameter further comprises: horizontal surface solar radiation per unit time;

[0075] The power generation in the future preset time is calculated by the following formula: E=H×P×K×T;

[0076] Wherein, E is the power generation in the future preset time, H is the horizontal surface solar radiation per unit time, P is the system installed capacity, K is the comprehensive efficiency coefficient, and T is the future preset time.

[0077] According to an embodiment of the present disclosure, the power surplus or deficit of the any micro-grid in the future preset time is determined according to the current energy storage capacity in the any micro-grid, the power generation in the future preset time, and the power consumption in the future preset time, comprising:

[0078] The power surplus or deficit in the future preset time is calculated by the following formula: Q=A+B-C;

[0079] Wherein, Q is the power surplus or deficit in the future preset time, A is the current energy storage capacity in the any micro-grid, B is the power generation in the future preset time, and C is the power consumption in the future preset time.

[0080] When the Q is positive, the power surplus or deficit in the future preset time is used to indicate that the any micro-grid is in a power surplus state; when the Q is negative, the power surplus or deficit in the future preset time is used to indicate that the any micro-grid is in a power deficit state, and when the Q is zero, the power surplus or deficit in the future preset time is used to indicate that the any micro-grid is in a power balance state.

[0081] According to an embodiment of the present disclosure, the latest current electricity price of the any micro-grid is obtained according to the power generation cost, the transmission cost and the active power loss of the any micro-grid, comprising:

[0082] The power generation cost of the any micro-grid is calculated based on the active power of the any micro-grid in a specified time, the current electricity price, the total discharge power of the any micro-grid, and the demand side response power consumption of the any micro-grid.

[0083] determining the transmission cost based on the active power from the any micro-grid to the transmission line within the specified time;

[0084] determining the active power loss within the specified time based on the total transmission line topology level, the transmission line resistance and the transmission line current;

[0085] the latest current price is the price that satisfies the power generation cost less than a first preset constant, the active power loss less than a second preset constant and the transmission cost minimum.

[0086] In a third aspect, the embodiments of the present disclosure provide an intelligent source-grid-load-storage coordinated power consumption regulation method, applied to an energy regulation center, wherein the energy regulation center is in communication connection with an energy regulation unit, the energy regulation unit is arranged in each micro-grid in a micro-grid cluster, the energy regulation units in each micro-grid in the micro-grid cluster are in communication connection, and the intelligent regulation method comprises:

[0087] receiving the power consumption surplus / deficit amount within a future preset time sent by the energy regulation unit;

[0088] receiving the latest current price sent by the energy regulation unit;

[0089] generating a micro-grid coordinated power consumption regulation scheme using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power consumption surplus / deficit amount within the future preset time and the latest current price of each micro-grid in the micro-grid cluster;

[0090] sending the micro-grid coordinated power consumption regulation scheme to the energy regulation unit in any micro-grid.

[0091] According to the embodiments of the present disclosure, the generation of the micro-grid coordinated power consumption regulation scheme using the fusion multi-objective algorithm based on the Pareto frontier curve and the fuzzy algorithm according to the power consumption surplus / deficit amount within the future preset time and the latest current price of each micro-grid in the micro-grid cluster comprises:

[0092] fuzzy processing the power consumption surplus / deficit amount and the latest current price using the fuzzy algorithm; and converting the power consumption surplus / deficit amount and the latest current price after the fuzzy processing into membership values in a fuzzy set through a pre-defined membership function;

[0093] allocating corresponding weights to a plurality of optimization objectives according to preset parameters; the optimization objectives include minimizing power consumption cost, maximizing energy utilization efficiency or reducing environmental impact;

[0094] According to the multiple optimization objectives and the weights corresponding to each optimization objective, a multi-objective optimization algorithm is used to perform iterative calculation on the membership values in the fuzzy set, to generate multiple preliminary power regulation schemes, the multi-objective optimization algorithm including a particle swarm optimization algorithm or a genetic algorithm;

[0095] The multiple objective function values of each preliminary power regulation scheme are evaluated, and the positions of each preliminary power regulation scheme are marked on the Pareto frontier curve;

[0096] The micro-grid collaborative power regulation scheme is selected or generated by a preset fusion strategy in combination with the Pareto frontier curve and the membership values in the fuzzy set, the preset fusion strategy including weighted fusion.

[0097] In a fourth aspect, the disclosure provides a source-network-load-storage collaborative power intelligent regulation device, which is arranged in an energy regulation unit, the energy regulation unit is arranged in each micro-grid in a micro-grid cluster, and is in communication connection with an energy regulation center in the micro-grid cluster, the energy regulation units in each micro-grid in the micro-grid cluster are in communication connection, and the intelligent regulation device in the energy regulation unit in any micro-grid includes:

[0098] A parameter acquisition module is arranged to acquire historical power consumption information of the any micro-grid and meteorological parameters in a future preset time, the historical power consumption information including historical electrical load and historical thermal load, and the meteorological parameters including one or more of the following parameters: illumination intensity, temperature, wind speed, wind direction, and relative humidity;

[0099] The power consumption prediction module is configured to jointly predict the power consumption of the any micro-grid in the future preset time according to the historical power consumption information and the meteorological parameters by using a LRCN double-layer network combination model based on a time attention mechanism; the LRCN double-layer network combination model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module and a multi-source data feature sharing layer module; the joint prediction of the power consumption of the any micro-grid in the future preset time according to the historical power consumption information and the meteorological parameters by using the LRCN double-layer network combination model based on the time attention mechanism comprises: processing the historical electric load by using a first load feature extraction channel to obtain a feature representation of the historical electric load; processing the historical thermal load by using a second load feature extraction channel to obtain a feature representation of the historical thermal load; processing the temperature in the meteorological parameters, the feature representation of the historical electric load and the feature representation of the historical thermal load by using the multi-source data feature sharing layer module to obtain an electric load prediction value and a thermal load prediction value of the any micro-grid in the future preset time; and obtaining the power consumption of the any micro-grid in the future preset time according to the electric load prediction value and the thermal load prediction value.

[0100] The power consumption surplus or deficit calculation module is configured to determine the power generation of the any micro-grid in the future preset time according to the meteorological parameters; and determine the power consumption surplus or deficit of the any micro-grid in the future preset time according to the current energy storage capacity in the any micro-grid, the power generation in the future preset time and the power consumption in the future preset time.

[0101] The latest current electricity price calculation module is configured to obtain the latest current electricity price of the any micro-grid according to the power generation cost, the transmission cost and the active power loss of the any micro-grid.

[0102] The parameter sending module is configured to send the power consumption surplus or deficit in the future preset time to the energy regulation and control center; and send the latest current electricity price to the energy regulation and control center.

[0103] The regulation scheme receiving module is configured to receive a micro-grid coordinated power consumption regulation scheme returned by the energy regulation and control center.

[0104] The regulation scheme executing module is configured to perform power consumption regulation according to the micro-grid coordinated power consumption regulation scheme.

[0105] In a fifth aspect, the disclosure provides an intelligent power consumption regulation device for source-network-load coordination, which is arranged in an energy regulation center. The energy regulation center is in communication connection with an energy regulation unit. The energy regulation unit is arranged in each micro-grid in a micro-grid cluster. The energy regulation units in each micro-grid in the micro-grid cluster are in communication connection. The intelligent regulation device comprises:

[0106] a parameter receiving module arranged to receive power consumption surplus / deficit in a future preset time sent by the energy regulation unit, and receive the latest current price sent by the energy regulation unit;

[0107] a regulation scheme generating module arranged to generate a micro-grid coordinated power consumption regulation scheme using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power consumption surplus / deficit in the future preset time and the latest current price of each micro-grid in the micro-grid cluster;

[0108] a regulation scheme sending module arranged to send the micro-grid coordinated power consumption regulation scheme to the energy regulation unit in any micro-grid.

[0109] In a sixth aspect, the disclosure provides an electronic device comprising a memory and a processor. The memory is configured to store one or more computer instructions. The one or more computer instructions are executed by the processor to implement the intelligent regulation method of any one of the second aspect and the third aspect.

[0110] In a seventh aspect, the disclosure provides a computer readable storage medium having computer instructions stored thereon. The computer instructions are executed by a processor to implement the intelligent regulation method of any one of the second aspect and the third aspect.

[0111] In an eighth aspect, the disclosure provides a computer program product comprising a computer program. The computer program is executed by a processor to implement the intelligent regulation method of any one of the second aspect and the third aspect.

[0112] According to the technical scheme provided by the embodiment of the present disclosure, an intelligent source-grid-load-storage collaborative power consumption regulation system is provided, which comprises an energy regulation center and an energy regulation unit arranged in a micro-grid. Each energy regulation unit is responsible for collecting historical power consumption information of the micro-grid, meteorological parameters and other data in a future preset time, and using a LRCN double-layer network combination model based on a time attention mechanism to predict power consumption. At the same time, each energy regulation unit is also responsible for determining the power generation capacity, energy storage capacity and power consumption surplus / deficit in the future preset time of the micro-grid, and sending relevant information to the energy regulation center. The energy regulation center uses a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm to generate a micro-grid collaborative power consumption regulation scheme according to the power consumption surplus / deficit and the latest current electricity price and other information of each micro-grid, and sends the regulation scheme to each energy regulation unit for execution, ensuring the energy supply and demand balance of the micro-grid. Thus, efficient, intelligent and fine management of energy in the micro-grid cluster is realized, thereby improving energy utilization efficiency, reducing energy consumption and cost, and providing strong support for the sustainable development of the micro-grid.

[0113] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0114] Other features, objects and advantages of the present disclosure will become more apparent from the following detailed description of the non-limiting embodiments, combined with the attached drawings. In the drawings:

[0115] Fig. 1 shows a structural schematic diagram of an intelligent source-grid-load-storage collaborative power consumption regulation system according to an embodiment of the present disclosure;

[0116] Fig. 2 shows a schematic diagram of processing historical electrical load and historical thermal load and temperature using a LRCN double-layer network combination model based on a time attention mechanism according to an embodiment of the present disclosure;

[0117] Fig. 3 shows a flowchart of an intelligent source-grid-load-storage collaborative power consumption regulation method according to an embodiment of the present disclosure;

[0118] Fig. 4 shows a flowchart of a method for calculating power consumption in a future preset time according to an embodiment of the present disclosure;

[0119] Fig. 5 shows a flowchart of a method for optimizing a LRCN double-layer network combination model based on a time attention mechanism according to an embodiment of the present disclosure;

[0120] Fig. 6 shows a flowchart of another intelligent source-grid-load-storage collaborative power consumption regulation method according to an embodiment of the present disclosure;

[0121] FIG. 7 shows a flow chart of a method for generating a micro-grid coordinated power consumption regulation scheme based on a Pareto frontier curve and a fuzzy algorithm according to an embodiment of the present disclosure;

[0122] FIG. 8 shows a structural block diagram of a source-grid-load-storage coordinated power consumption intelligent regulation device according to an embodiment of the present disclosure;

[0123] FIG. 9 shows a structural block diagram of another source-grid-load-storage coordinated power consumption intelligent regulation device according to an embodiment of the present disclosure;

[0124] FIG. 10 shows a structural block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0125] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so as to be easily implemented by those skilled in the art. Also, parts irrelevant to the description of the exemplary embodiments are omitted in the accompanying drawings for the sake of clarity.

[0126] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate that there are features, numbers, steps, actions, components, parts or combinations thereof disclosed in the specification, and do not exclude the possibility of one or more other features, numbers, steps, actions, components, parts or combinations thereof existing or added.

[0127] It is also necessary to note that the embodiments in the present disclosure and the features in the embodiments can be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0128] As described above, the energy interaction and coordination among the micro-grids in the prior art are insufficient, resulting in low energy utilization efficiency, increased operation cost and risk, and other problems.

[0129] In recent years, source-grid-load-storage is a new operation mode in the power system, which is planned as a whole with "power source, power grid, load, and energy storage". Among them, "source" refers to traditional thermal power, hydropower, and various power generation methods such as wind power and solar power; "grid" refers to the power transmission and distribution grid in the power system, as well as the application of smart grid technology; "load" refers to various users in the power system, including industry, commerce, and residents; and "storage" mainly refers to battery, super capacitor, and other energy storage devices.

[0130] To solve the above problems in the prior art, the present disclosure proposes a source network load storage collaborative electricity use intelligent regulation and control system. The system integrates advanced energy management algorithms and intelligent control strategies, and fuses source network load storage based on big data, artificial intelligence and other technologies for collaborative electricity use deployment. By introducing advanced deep learning technology, the prediction accuracy of microgrid energy supply and demand is improved. At the same time, through the fusion of Pareto frontier curve and fuzzy algorithm based multi-objective algorithm, the collaborative optimization of each microgrid in the microgrid cluster is realized. In addition, the system also has flexible regulation and control strategies, which can quickly respond and adjust according to real-time energy supply and demand, ensuring the energy supply and demand balance and efficient use of microgrids, and realizing efficient, intelligent and fine management of energy in the microgrid cluster, thereby improving energy utilization efficiency and reducing energy consumption and cost.

[0131] To support a complete process from generating data for source network load storage collaborative electricity use intelligent regulation and control to formulating source network load storage collaborative electricity use intelligent regulation and control strategies according to the data and finally using the intelligent regulation and control strategies for electricity regulation and control of each microgrid, the present disclosure introduces an energy regulation and control unit and an energy regulation and control center. The energy regulation and control unit is arranged in each microgrid in the microgrid cluster and can be a separate device or a functional module integrated in the existing device of each microgrid. The energy regulation and control center can establish a communication connection with the energy regulation and control unit in each microgrid. Through the established communication connection, the energy regulation and control center can receive energy-related data collected by the energy regulation and control unit in each microgrid in real time, and use advanced algorithms and models to analyze the data to form scientific and reasonable energy regulation and control strategies. At the same time, the energy regulation and control center can also send the formed energy regulation and control strategies back to the corresponding energy regulation and control unit to guide it to deploy and manage energy. In this way, the energy regulation and control unit and the energy regulation and control center cooperate to realize the source network load storage collaborative electricity use intelligent regulation and control.

[0132] In the embodiments of the present disclosure, the energy regulation and control unit and the energy regulation and control center can be one or more processors, controllers or chips with communication interfaces that can implement communication protocols, and can also include memories and related interfaces, system transmission buses, etc. if necessary; the processor, controller or chip executes program-related code to realize corresponding functions.

[0133] FIG. 1 shows a structural schematic diagram of a source-network-load-storage coordination intelligent regulation system according to an embodiment of the present disclosure. As shown in FIG. 1, the intelligent regulation system comprises an energy regulation center and an energy regulation unit arranged in each micro-grid in a micro-grid cluster, the energy regulation center is in communication connection with the energy regulation unit in each micro-grid (two energy regulation units are taken as an example for illustration in FIG. 1), and the energy regulation units in each micro-grid in the micro-grid cluster are in communication connection. The micro-grid cluster is a network composed of multiple micro-grids through certain connection methods (such as power lines, communication lines, etc.), each micro-grid is a regional power system with autonomous control, protection and management functions, and can provide reliable, efficient and clean electric energy for users. Each micro-grid comprises distributed energy (such as solar energy, wind energy), energy storage devices, loads, etc., and is connected to the main grid through a tie line. Due to the power uncertainty of distributed power generation and loads, they can realize bidirectional energy exchange with the main grid.

[0134] According to an embodiment of the present disclosure, the energy regulation unit in any micro-grid in the micro-grid cluster is configured to: acquire historical power consumption information of the any micro-grid and meteorological parameters in a future preset time; jointly predict power consumption of the any micro-grid in the future preset time according to the historical power consumption information and the meteorological parameters using a LRCN double-layer network combination model based on a time attention mechanism; determine power generation of the any micro-grid in the future preset time according to the meteorological parameters; determine power surplus / deficit of the any micro-grid in the future preset time according to current energy storage capacity in the any micro-grid and the power generation in the future preset time and the power consumption in the future preset time; send the power surplus / deficit in the future preset time to the energy regulation center; acquire the latest current price of the any micro-grid according to generation cost, transmission cost and active power loss of the any micro-grid; and send the latest current price to the energy regulation center.

[0135] According to an embodiment of the present disclosure, the energy regulation center is configured to generate a micro-grid coordinated power consumption regulation scheme using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power surplus / deficit of each micro-grid in the micro-grid cluster in the future preset time and the latest current price; and send the micro-grid coordinated power consumption regulation scheme to the energy regulation unit in the any micro-grid.

[0136] According to an embodiment of the present disclosure, the energy regulation unit is further configured to perform power consumption regulation according to the micro-grid coordinated power consumption regulation scheme.

[0137] The specific functions of the energy regulation units in each micro-grid and the energy regulation center in the source-grid-load-storage collaborative intelligent power regulation system are described in detail below.

[0138] For the energy regulation unit in each micro-grid in the micro-grid cluster, the energy regulation unit is mainly responsible for power consumption prediction, power generation determination, power consumption surplus or deficit calculation, and power price updating within the micro-grid, as well as communication with the energy regulation center.

[0139] According to an embodiment of the present disclosure, the historical power consumption information includes but is not limited to historical electrical load and historical thermal load.

[0140] The electrical load refers to the power consumption of all electrical equipment in the micro-grid. The historical electrical load refers to the electrical load data record of the micro-grid at a specific time point within a certain time period in the past. These data record the actual power mode of the power user at a certain time, providing a basis for energy management and optimization.

[0141] The thermal load refers to the energy consumption of equipment that needs to be heated or cooled in the micro-grid. The historical thermal load refers to the record data of the heat or thermal power transferred by the system or equipment in the micro-grid per unit time, as well as the thermal load demand of the entire micro-grid in different time periods within a certain time period in the past. These data help to analyze the thermal demand and energy saving potential of the micro-grid, providing guidance for the optimization of the thermal system.

[0142] The present disclosure considers the historical data of electrical load and thermal load, which can more comprehensively understand the operation of the micro-grid and the load variation law. This helps to more accurately predict the power demand in the future, providing a scientific basis for the energy planning and power dispatching of the micro-grid. In addition, meteorological parameters such as temperature, humidity, and wind speed are introduced, which can further consider the influence of weather changes on power demand. Especially for air conditioning, heating and other loads that are greatly affected by weather, the introduction of meteorological parameters can significantly improve the accuracy of prediction. Based on the high-precision prediction results, the micro-grid can more accurately arrange the production and dispatching of energy. For example, when high-temperature weather is predicted, the output of photovoltaic power generation can be increased in advance, and the strategy of energy storage devices can be adjusted to meet the growth of power load. Therefore, through the multi-source type analysis method based on historical electrical load and historical thermal load and meteorological parameters, the prediction of future electrical load can be more accurate due to the consideration of the correlation between electrical, thermal and weather during the actual energy demand process.

[0143] According to an embodiment of the present disclosure, the energy regulation unit obtains the historical power consumption information of any micro-grid in the following ways:

[0144] The first way is that the energy regulation unit obtains historical electricity information from smart meters and sensors in the microgrid. Smart meters are responsible for real-time measurement and recording of electrical load data, including voltage, current, power factor, and electricity consumption data such as daily and monthly electricity consumption. For the collection of thermal load data, sensors installed on thermal load equipment (such as boilers, heaters, etc.) can measure and record temperature, heat flow, and other parameters. These sensors can monitor the operating status of thermal load equipment in real time. In this way, smart meters and sensors can record and transmit electricity and heat data in real time, and send historical electricity information to the energy regulation unit for analysis and processing.

[0145] The second way is that in the case of thermal load equipment (such as boilers, heaters) and electrical load equipment (such as motors, lighting systems) in the microgrid having data output function, the energy regulation unit can directly read the historical electricity information of these devices.

[0146] The third way is that in some cases, the microgrid may share data with external energy suppliers or service providers. The energy regulation unit can obtain historical electricity information containing thermal load and electrical load from these third-party data sources through a secure data exchange mechanism.

[0147] The fourth way is to equip each microgrid with a data management system (DMS) for storing, managing, and maintaining data related to energy use, electrical load, and thermal load. When obtaining historical electricity information of any microgrid, the energy regulation unit can interface with the DMS to periodically or in real time obtain historical electricity information containing thermal load and electrical load from the database, so that the energy regulation unit can further process and analyze the data, such as calculating electricity consumption trends and predicting electricity demand, to provide decision support for energy regulation.

[0148] In order to protect the security and privacy of data when obtaining historical electricity information, appropriate data encryption and access control measures can be taken. Through the analysis of historical electricity information, the energy regulation unit can optimize the energy management strategy of the microgrid, improve energy utilization efficiency, and reduce energy consumption and cost.

[0149] According to an embodiment of the present disclosure, the energy regulation unit is further configured to obtain meteorological parameters in a future preset time. The "future preset time" refers to a preset time between the current time and a future time, or a preset time between a future time and another future time, such as a week or a month in the future, and is set according to design requirements. The meteorological parameters include, but are not limited to, one or more of the following parameters: light intensity, temperature, wind speed, wind direction, and relative humidity. Among them, temperature is one of the main factors affecting power load. The increase in temperature will lead to an increase in demand for air conditioners and refrigeration equipment, thereby increasing power load. For example, in hot summer, people use air conditioners and other power equipment to cool down, resulting in an increase in power demand.

[0150] In order to obtain meteorological parameters, the energy regulation unit can be configured to establish interface communication with reliable meteorological service providers, such as the National Weather Service and commercial meteorological data providers. Meteorological data is accessed through an API (Application Program Interface), which usually provides historical and predicted meteorological data.

[0151] In obtaining meteorological parameters in a future preset time, the energy regulation unit first sets the time range of the requested meteorological data (including the time period in the future preset time) and selects the required meteorological parameters to obtain the meteorological parameters in the future preset time. Then, the API sends a request including the time range and the required meteorological parameters to the meteorological service application, and receives the corresponding meteorological data returned by the meteorological service application, usually in the form of a JSON or XML data packet. Finally, the energy regulation unit parses the received meteorological data and extracts the required meteorological parameter values.

[0152] When designing a learning model for predicting electricity consumption in a future preset time based on the obtained historical electrical load, historical thermal load, and meteorological parameters, the present inventors have conducted careful research and verification, and selected a LRCN (Long-term Recurrent Convolutional Network) double-layer network combination model combining convolutional neural network (CNN) and recurrent neural network (RNN), and introduced an efficient multi-scale attention module (Efficient Multi-Scale Attention Module) based on the model, forming a LRCN double-layer network combination model based on time attention mechanism.

[0153] The LRCN double-layer network combination model based on the time attention mechanism combines the characteristics of the convolutional neural network (CNN) and the recurrent neural network (RNN), specifically: CNN can capture local spatial features, which is very effective when processing meteorological parameters such as the spatial distribution of temperature, humidity, etc. RNN, especially LSTM or GRU, is good at processing time series data and can capture long-term dependencies of power consumption over time.

[0154] In the specific construction of the model, for the convolutional neural network (CNN), the disclosure designs multiple convolutional layers, each of which contains multiple convolution kernels (filters) for extracting local features of the input data. CNN captures these features by sliding its convolution kernels over the data and performing convolution operations. After each convolutional layer, an activation function (such as ReLU) is used to increase the nonlinearity of the model, and a pooling layer (such as max pooling) is added as needed to reduce the spatial dimension and preserve key information. In addition, after multiple convolutional layers and pooling layers, a flattening layer is used to convert the multi-dimensional feature map into a one-dimensional vector for input to the fully connected layer; for the long short-term memory network (LSTM) module, one or more LSTM layers are designed, each of which contains multiple LSTM units. LSTM units use a gating mechanism (forget gate, input gate, and output gate) to remember, forget, and update information. The output of the LSTM layer can be used as the input of the next LSTM layer or input to the fully connected layer for further processing; for the efficient multi-scale attention module, a learnable weight matrix and an activation function (such as softmax) are used to calculate the attention score, so that the weight can be assigned at different time steps and different feature dimensions, and the attention score is applied to the output of the LSTM layer or the input of the fully connected layer, so that the model can focus on key information.

[0155] In addition, the present inventors fully consider the characteristics of historical electrical load and historical thermal load, i.e., for electrical load, the electrical load has the characteristics of time-varying, that is, the electrical load changes over time and presents obvious daily, monthly and annual periodic changes. The load curve in a day usually has two peaks, in the morning and evening respectively, and the efficient multi-scale attention module can capture this time-varying by focusing on features of different time scales (such as hours, days, weeks, months, etc.) to better understand and predict the trend of electrical load changes; the electrical load also has the characteristics of uncertainty, that is, the change of the electrical load is affected by many factors such as temperature, season, economic situation, etc., and these factors have uncertainty, and the multi-scale attention module can consider these uncertain factors at multiple scales by assigning different attention weights to highlight the influence of important factors, thereby improving the accuracy of prediction; in addition, the electrical load also has the characteristics of volatility, and the efficient multi-scale attention module can focus on different scale performances of this volatility to more accurately predict the instantaneous changes of the electrical load. Similarly, for thermal load, the thermal load has the characteristics of large peak-valley load difference, that is, the thermal load usually has significant peak-valley difference during the heating period, which is related to factors such as temperature, heating equipment and building structure, and the efficient multi-scale attention module can allocate attention at different time scales for this peak-valley difference to more accurately predict the peak and trough periods of thermal load; in addition, the thermal load also has the characteristics of poor stability, that is, the thermal load is affected by many factors and is prone to change, and through the efficient multi-scale attention module, the model can capture these influencing factors at multiple scales and learn how to adapt to these changes to improve the stability of prediction.

[0156] In summary, the efficient multi-scale attention module is introduced into the model, which enables the model to consider information of different time scales at the same time, thereby capturing the characteristics of historical electricity consumption information and meteorological parameters at different scales. This multi-scale feature extraction capability helps the model to more comprehensively and deeply understand the internal laws and trends of the data. Moreover, it can further enhance the effect of the time attention mechanism, enabling the model to more accurately identify and focus on time points and features closely related to future electricity consumption changes, thereby more accurately capturing key information in historical electricity consumption information and meteorological parameters and more accurately predicting future electricity consumption to provide more reliable data support for energy regulation. Finally, by adjusting the parameters and configurations of the multi-scale attention module, the model can adapt to different prediction scenarios and data sets, thereby realizing more personalized prediction services.

[0157] According to an embodiment of the present disclosure, the LRCN double-layer network combination model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module, and a multi-source data feature sharing layer module.

[0158] In an embodiment of the present disclosure, the CNN module, the LSTM module, the efficient multi-scale attention module, and the multi-source data feature sharing layer module can each be one or more processors, controllers, or chips with a communication interface capable of implementing a communication protocol, and can further include a memory and related interfaces, a system transmission bus, and the like, if necessary; the processors, controllers, or chips execute program-related codes to implement corresponding functions. Alternatively, the CNN module, the LSTM module, the efficient multi-scale attention module, and the multi-source data feature sharing layer module can share an integrated chip or share processors, controllers, memories, and the like. The shared processors, controllers, or chips execute program-related codes to implement corresponding functions.

[0159] FIG. 2 shows a schematic diagram of processing historical electrical load and historical thermal load and temperature using the LRCN double-layer network combination model based on the time attention mechanism according to an embodiment of the present disclosure. The LRCN double-layer network combination model based on the time attention mechanism is used to jointly predict the electrical consumption of any micro-grid in a future preset time period according to historical electricity consumption information and meteorological parameters, which comprises: processing the historical electrical load using a first load feature extraction channel to obtain a feature representation of the historical electrical load; processing the historical thermal load using a second load feature extraction channel to obtain a feature representation of the historical thermal load; processing the temperature in the meteorological parameters, the feature representation of the historical electrical load, and the feature representation of the historical thermal load using the multi-source data feature sharing layer module to obtain electrical load prediction values and thermal load prediction values of any micro-grid in the future preset time period; and obtaining the electrical consumption of any micro-grid in the future preset time period according to the electrical load prediction values and the thermal load prediction values.

[0160] The present disclosure uses the load feature extraction channel comprising the CNN module, the LSTM module, and the efficient multi-scale attention module to respectively extract features of the historical electrical load and the historical thermal load, and respectively obtain feature representations of the historical electrical load and the historical thermal load.

[0161] Specifically, the first load feature extraction channel includes a first convolutional neural network (CNN) module, a first long short-term memory (LSTM) module, and a first efficient multi-scale attention module, the historical electrical load is processed using the first load feature extraction channel to obtain a feature representation of the historical electrical load, including:

[0162] The first CNN module is used to extract local features of the historical electrical load through convolution operation to obtain a feature map of the historical electrical load, the first LSTM module is used to process the feature map of the historical electrical load to capture long-term and short-term dependencies in the historical electrical load based on the characteristics of a recurrent connection to obtain a hidden state vector of the historical electrical load, and the first efficient multi-scale attention module is used to process the hidden state vector of the historical electrical load to obtain the feature representation of the historical electrical load weighted by an attention mechanism by assigning different weights to different features.

[0163] According to an embodiment of the present disclosure, the second load feature extraction channel includes a second CNN module, a second LSTM module, and a second efficient multi-scale attention module, the historical thermal load is processed using the second load feature extraction channel to obtain a feature representation of the historical thermal load, including: the second CNN module is used to extract local features of the historical thermal load through convolution operation to obtain a feature map of the historical thermal load, the second LSTM module is used to process the feature map of the historical thermal load to capture long-term and short-term dependencies in the historical thermal load based on the characteristics of a recurrent connection to obtain a hidden state vector of the historical thermal load, and the second efficient multi-scale attention module is used to process the hidden state vector of the historical thermal load to obtain the feature representation of the historical thermal load weighted by an attention mechanism by assigning different weights to different features.

[0164] The local features refer to short-term, local patterns or structures in the data. For example, for the electrical load data, the local features can include the electrical load fluctuations, peak values, valley values, etc. in a certain time period. The hidden state vector synthesizes the information in the historical electrical load data, and contains the representation of long-term and short-term dependencies. This vector can be regarded as a kind of high-level feature representation. The long-term and short-term dependencies refer to the relevance between different time points in the data. In the long-term dependency, the data value at a time point can be affected by the data value a long time ago; while in the short-term dependency, the data value at a time point can be mainly affected by the data values of its adjacent time points. In time series analysis, for example, for the electrical load data, the long-term dependency can reflect seasonal trends, periodic changes, etc., while the short-term dependency can reflect daily changes, weather influences, etc. In the efficient multi-scale attention module, different features (including the aforementioned local features, long-term and short-term dependencies, etc.) are given different weights. This weight reflects the importance or relevance of these features in predicting future electrical load or thermal load.

[0165] In the embodiments of the present application, the first convolutional neural network CNN module, the first long short-term memory network LSTM module, the first efficient multi-scale attention module, the second convolutional neural network CNN module, the second long short-term memory network LSTM module and the second efficient multi-scale attention module can be one or more processors, controllers or chips with a communication interface capable of implementing a communication protocol, and can also include a memory and related interfaces, a system transmission bus, etc. if necessary; the processor, controller or chip executes program-related codes to realize corresponding functions. Alternatively, the first convolutional neural network CNN module, the first long short-term memory network LSTM module, the first efficient multi-scale attention module, the second convolutional neural network CNN module, the second long short-term memory network LSTM module and the second efficient multi-scale attention module share an integrated chip or share a processor, controller, memory, etc. The shared processor, controller or chip executes program-related codes to realize corresponding functions.

[0166] In specific applications, a pre-trained LRCN double-layer network combination model based on the time attention mechanism can be used to perform forward propagation on the historical electrical load and thermal load data to extract features.

[0167] When training the model, the following steps can be performed:

[0168] Step 1, data preprocessing: collect and organize historical electricity consumption information (including historical electrical load and historical thermal load) and meteorological parameter data, and clean and organize them. Divide the data into training set, validation set and test set, usually the training set accounts for the majority, the validation set is used for model selection and parameter tuning, and the test set is used for final evaluation of model performance. According to the task requirements, standardize, normalize or other necessary preprocessing operations are performed on the data.

[0169] Step 2, model initialization: initialize the parameters of the LRCN double-layer network model, including the weights and biases of the convolutional neural network CNN module, the long short-term memory network LSTM module and the efficient multi-scale attention module. Initialize learning rate, batch size, iteration number and other hyperparameters.

[0170] Step 3, build LRCN double-layer network based on time attention mechanism: design an LRCN network structure containing CNN (for feature extraction) and LSTM (for processing time series information). Add an efficient multi-scale attention module after the LSTM layer to calculate the weights of different time step features and highlight important features.

[0171] Step 4, rolling mechanism (time window): at the beginning of training, select a starting time point and set the size of the time window, that is, the length of the data input to the model each time. At the same time, set the rolling step, that is, the number of time steps rolled forward each time. Starting from the starting time point, take the data within the time window as the input of the first training batch. In the training process, use the time window to capture the time series dependence in the historical data. Take the data within the time window as the input of the model and predict the electricity consumption within a certain preset time after the time window. As time goes on, the time window is constantly rolled forward, and the model is updated with new data.

[0172] Step 5, training loop: for each training batch, extract data within a time window from the training set. Input the data into the LRCN double-layer network model, calculate the predicted value through forward propagation. Calculate the loss function (such as mean square error MSE) between the predicted value and the true value. Calculate the gradient through the back propagation algorithm, and update the model parameters using the optimizer (such as Adam). After training each batch, roll the time window forward according to the rolling step and update the training data. If the end time of the time window exceeds the last time of the training set after rolling forward, you can choose one of the following methods to handle it:

[0173] Reset the starting time point to a time point before the starting point of the current time window and continue rolling. Or,

[0174] Use a circular way to set the starting point of the time window back to the starting time point of the training set, realizing the circular use of data.

[0175] Repeat step 5 above until a preset number of iterations is reached or performance on the validation set no longer improves.

[0176] Step 6, verification and testing: Evaluate the performance of the model on the validation set, and make parameter adjustments and network structure adjustments as needed. Use the test set to evaluate the final model, calculate prediction error, accuracy and other indicators.

[0177] Step 7, model saving and deployment: Save the trained model parameters and structure. Deploy the model to the actual environment and use new data for prediction.

[0178] Through the above training steps 1-7 and the use of the rolling mechanism, the historical electricity consumption information and meteorological parameters can be effectively used to train the LRCN double-layer network combination model based on the time attention mechanism, and the prediction performance can be improved.

[0179] In addition, in order to realize adaptive prediction of changing load, ensure that the model can capture the feature changes of the system through rolling update, and complete accurate prediction of the load, the disclosure also uses an online learning framework based on transfer learning to optimize the model.

[0180] According to an embodiment of the disclosure, the energy regulation unit is further configured to: obtain a current actual electrical load and a current actual thermal load; obtain an electrical load deviation according to the current actual electrical load and the electrical load prediction value; obtain a thermal load deviation according to the current actual thermal load and the thermal load prediction value; when the electrical load deviation and / or the thermal load deviation do not meet a preset requirement, perform parameter transfer on the LRCN double-layer network combination model based on the time attention mechanism using an online learning framework based on transfer learning, including: if a difference between an average value of the historical electrical load and the current actual electrical load is greater than a preset electrical load difference threshold, updating the first load feature extraction channel parameter; if a difference between an average value of the historical thermal load and the current actual thermal load is greater than a preset thermal load difference threshold, updating the second load feature extraction channel parameter; obtaining a current temperature and a historical temperature, and if a difference between an average value of the historical temperature and the current temperature is greater than a preset temperature difference threshold, updating the first load feature extraction channel parameter and the second load feature extraction channel parameter.

[0181] When updating the first load feature extraction channel parameter and / or the second load feature extraction channel parameter, it can be specifically performed by fixing the model bottom layer network parameter and fine-tuning the top layer network parameter, specifically as follows:

[0182] First, identify the bottom and top networks in the first and second load feature extraction channels. Typically, the bottom network includes layers closer to the input (e.g., first convolutional layers), while the top network includes layers closer to the output (e.g., fully connected layers or LSTM layers).

[0183] Then, when fine-tuning the model, set the bottom network parameters of the first and second load feature extraction channels as untrainable. This can be achieved by setting the parameters' requires_grad attribute to False (in PyTorch) or using the corresponding framework's method to freeze layer parameters.

[0184] Finally, when updating the first load feature extraction channel parameters (i.e., the electrical load difference exceeds the threshold), ensure that only the top network parameters of the first load feature extraction channel are fine-tuned. This can be achieved by setting the requires_grad attribute of these parameters to True (in PyTorch) or allowing these layer parameters to update during the optimization process; when updating the second load feature extraction channel parameters (i.e., the thermal load difference exceeds the threshold), similarly, only the top network parameters of the second load feature extraction channel are fine-tuned.

[0185] The present disclosure takes into account that temperature changes may affect the prediction of both electrical and thermal loads, so if the difference between the average value of historical temperature and the current temperature is greater than the preset temperature difference threshold, the top network parameters of both channels need to be updated.

[0186] During the fine-tuning process, appropriate loss functions (such as mean square error, cross-entropy, etc.) and optimizers (such as SGD, Adam, etc.) can be used to update the top network parameters. Ensure that during training or fine-tuning, the bottom network parameters remain unchanged, while the top network parameters are updated according to the backpropagation algorithm.

[0187] After updating the parameters, use the validation set or test set to evaluate the performance of the model. By comparing the performance before and after updating, you can determine whether the parameter update is effective.

[0188] In addition, since load and temperature data may change over time, it is necessary to continuously monitor the performance of the model and make parameter updates or model adjustments as needed. The performance of the model can be re-evaluated periodically, and the preset threshold or update strategy can be adjusted as needed.

[0189] According to an embodiment of the present disclosure, the meteorological parameters further include: the horizontal surface solar radiation per unit time.

[0190] The power generation in the future preset time is calculated by the following formula: E = H x P x K x T;

[0191] Wherein, E is the power generation in the future preset time, H is the horizontal surface solar radiation in the unit time, P is the system installation capacity, K is the comprehensive efficiency coefficient, and T is the future preset time.

[0192] According to the embodiment of the present disclosure, the power surplus / deficit of the any micro-grid in the future preset time is determined according to the current energy storage capacity in the any micro-grid and the power generation in the future preset time and the power consumption in the future preset time, comprising:

[0193] The power surplus / deficit in the future preset time is calculated by the following formula, Q=A+B-C;

[0194] Wherein, Q is the power surplus / deficit in the future preset time, A is the current energy storage capacity in the any micro-grid, B is the power generation in the future preset time, and C is the power consumption in the future preset time.

[0195] When Q is positive, the power surplus / deficit in the future preset time is used to indicate that the any micro-grid is in the power surplus state; when Q is negative, the power surplus / deficit in the future preset time is used to indicate that the any micro-grid is in the power deficit state, and when Q is zero, the power surplus / deficit in the future preset time is used to indicate that the any micro-grid is in the power balance state.

[0196] According to the embodiment of the present disclosure, the latest current price of the any micro-grid is obtained according to the power generation cost, the transmission cost and the active power loss of the any micro-grid, comprising:

[0197] The power generation cost of the any micro-grid is calculated based on the active power of the any micro-grid, the current price, the total discharge power of the any micro-grid and the demand side response power consumption of the any micro-grid in a specified time; the transmission cost is determined based on the active power from the any micro-grid to the transmission line in the specified time; the active power loss in the specified time is determined based on the total transmission line topology level, the transmission line resistance and the transmission line current; the latest current price is the price that meets the conditions that the power generation cost is less than a first preset constant, the active power loss is less than a second preset constant and the transmission cost is the minimum.

[0198] For the energy regulation center, the energy regulation center is mainly responsible for generating a coordinated power consumption regulation scheme according to the data of each micro-grid and sending it to each micro-grid for execution.

[0199] According to an embodiment of the present disclosure, the power consumption surplus and deficiency of each micro-grid in the micro-grid cluster within the future preset time and the latest current electricity price are used to generate a micro-grid cooperative power consumption regulation scheme using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm, including: using the fuzzy algorithm to fuzz the power consumption surplus and deficiency and the latest current electricity price; converting the power consumption surplus and deficiency and the latest current electricity price after fuzzy processing into membership values in a fuzzy set through a predefined membership function; assigning corresponding weights to a plurality of optimization objectives according to preset parameters; the optimization objectives include: minimizing power consumption cost, maximizing energy utilization efficiency, or reducing environmental impact; according to the plurality of optimization objectives and the weights corresponding to each optimization objective, using a multi-objective optimization algorithm to iteratively calculate the membership values in the fuzzy set, generating a plurality of preliminary power consumption regulation schemes, the multi-objective optimization algorithm includes: particle swarm optimization algorithm or genetic algorithm; evaluating a plurality of objective function values of each preliminary power consumption regulation scheme, and marking the positions of each preliminary power consumption regulation scheme on the Pareto frontier curve; combining the Pareto frontier curve and the membership values in the fuzzy set, and selecting or generating the micro-grid cooperative power consumption regulation scheme through a preset fusion strategy; the preset fusion strategy includes: weighted fusion.

[0200] Figure 3 shows a source network load storage cooperative power consumption intelligent regulation method flow chart according to an embodiment of the present disclosure. The source network load storage cooperative power consumption intelligent regulation method is applied to an energy regulation unit, which is arranged in each micro-grid in a micro-grid cluster, and is in communication connection with an energy regulation center in the micro-grid cluster. The energy regulation units in each micro-grid in the micro-grid cluster are in communication connection, as shown in Figure 3, the source network load storage cooperative power consumption intelligent regulation method includes the following steps S310-S370:

[0201] In step S310, the energy regulation unit in any micro-grid in the micro-grid cluster obtains the historical power consumption information of the micro-grid and the meteorological parameters within the future preset time.

[0202] The historical power consumption information includes historical electrical load and historical thermal load, and the meteorological parameters include one or more of the following parameters: light intensity, temperature, wind speed, wind direction, and relative humidity.

[0203] In step S320, the historical power consumption information and the meteorological parameters are used to jointly predict the power consumption of the micro-grid within the future preset time using a LRCN double-layer network combination model based on a time attention mechanism.

[0204] The LRCN double-layer network combination model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module, and a multi-source data feature sharing layer module.

[0205] FIG. 4 shows a flowchart of a method for calculating power consumption in a future preset time according to an embodiment of the present disclosure. The power consumption in the future preset time of any micro-grid is jointly predicted using an LRCN double-layer network combination model based on a time attention mechanism according to the historical power consumption information and the meteorological parameters, as shown in FIG. 4, comprising the following steps S321-S324:

[0206] In step S321, the historical electric load is processed using a first load feature extraction channel to obtain a feature representation of the historical electric load.

[0207] Specifically, the first load feature extraction channel comprises a first convolutional neural network (CNN) module, a first long short-term memory (LSTM) module, and a first efficient multi-scale attention module. The historical electric load is processed using the first load feature extraction channel to obtain a feature representation of the historical electric load, which comprises: using the first CNN module to extract local features of the historical electric load through convolution operation to obtain a feature map of the historical electric load; using the first LSTM module to process the feature map of the historical electric load to capture long-term and short-term dependencies in the historical electric load based on the characteristics of the recurrent connection to obtain a hidden state vector of the historical electric load; and using the first efficient multi-scale attention module to process the hidden state vector of the historical electric load to obtain a feature representation of the historical electric load after attention mechanism weighting by assigning different weights to different features.

[0208] In step S322, the historical thermal load is processed using a second load feature extraction channel to obtain a feature representation of the historical thermal load.

[0209] Specifically, the second load feature extraction channel comprises a second convolutional neural network (CNN) module, a second long short-term memory (LSTM) module, and a second efficient multi-scale attention module, the historical thermal load is processed using the second load feature extraction channel to obtain a feature representation of the historical thermal load, which comprises: extracting local features of the historical thermal load by convolution operation using the second CNN module to obtain a feature map of the historical thermal load; processing the feature map of the historical thermal load using the second LSTM module to capture long-term and short-term dependencies in the historical thermal load based on the characteristics of the recurrent connection to obtain a hidden state vector of the historical thermal load; processing the hidden state vector of the historical thermal load using the second efficient multi-scale attention module to obtain a feature representation of the historical thermal load after attention mechanism weighting by assigning different weights to different features.

[0210] In step S323, the temperature in the meteorological parameter, the feature representation of the historical electrical load, and the feature representation of the historical thermal load are processed using the multi-source data feature sharing layer module to obtain an electrical load prediction value and a thermal load prediction value of the future preset time of any micro-grid.

[0211] In step S324, the electrical load prediction value and the thermal load prediction value are used to obtain the electricity consumption of the future preset time of any micro-grid.

[0212] In the implementation of this step S324, the thermal load prediction value can be first converted into an electrical energy consumption prediction value, and then the electrical load prediction value and the electrical energy consumption prediction value converted from the thermal load are superimposed to obtain a total electricity consumption prediction.

[0213] In step S330, the electricity generation of the future preset time of any micro-grid is determined according to the meteorological parameter; and the electricity surplus or deficit of the future preset time of any micro-grid is determined according to the current energy storage capacity in the micro-grid, the electricity generation of the future preset time, and the electricity consumption of the future preset time.

[0214] Specifically, the meteorological parameter further comprises a horizontal surface solar radiation amount per unit time, and the electricity generation of the future preset time is calculated by the following formula: E=H×P×K×T;

[0215] wherein E is the electricity generation of the future preset time, H is the horizontal surface solar radiation amount per unit time, P is the system installed capacity, K is the comprehensive efficiency coefficient, and T is the future preset time.

[0216] Specifically, the power surplus / deficit of the any micro-grid in the future preset time is determined according to the current energy storage capacity in the any micro-grid, the power generation in the future preset time, and the power consumption in the future preset time, and the power surplus / deficit of the any micro-grid in the future preset time comprises:

[0217] The power surplus / deficit in the future preset time is calculated by the following formula: Q=A+B-C.

[0218] Wherein, Q is the power surplus / deficit in the future preset time, A is the current energy storage capacity in the any micro-grid, B is the power generation in the future preset time, and C is the power consumption in the future preset time.

[0219] When the Q is positive, the power surplus / deficit in the future preset time indicates that the any micro-grid is in a power surplus state; when the Q is negative, the power surplus / deficit in the future preset time indicates that the any micro-grid is in a power deficit state, and when the Q is zero, the power surplus / deficit in the future preset time indicates that the any micro-grid is in a power balance state.

[0220] In step S340, the latest current electricity price of the any micro-grid is obtained according to the power generation cost, transmission cost and active power loss of the any micro-grid.

[0221] Specifically, first, the power generation cost of the any micro-grid is calculated based on the active power of the any micro-grid in a specified time, the current electricity price, the total discharge power of the any micro-grid, and the demand side response power consumption of the any micro-grid.

[0222] Wherein, the active power of the micro-grid refers to the actual AC power energy output or consumed by the power source (such as distributed generators, energy storage systems, renewable energy power generation equipment, etc.) in the micro-grid at a specific time, i.e. the energy actually converted or consumed.

[0223] The total discharge power of the micro-grid refers to the sum of the maximum power that all elements capable of discharging (such as energy storage devices, diesel generators, etc.) in the micro-grid can simultaneously output under certain conditions.

[0224] The demand side response (DR) power consumption of the micro-grid refers to the change in power consumption caused by the user actively adjusting its power consumption behavior according to the change in electricity price or grid guidance signal.

[0225] In specific implementation, the generation cost of any micro-grid can be calculated by multiplying the difference between the sum of active power of any micro-grid and demand-side response power consumption of any micro-grid within a specified time and the current electricity price after subtracting the total discharge power of any micro-grid, or can be calculated according to other ways, which can be set according to actual application requirements.

[0226] Then, the transmission cost is determined based on the active power from any micro-grid to the power transmission line within the specified time.

[0227] Among them, the active power of the micro-grid to the power transmission line refers to the electric power that can be converted into other forms of energy (such as mechanical energy, light energy, thermal energy, etc.) actually transmitted or exchanged by the micro-grid system through the power transmission line when connected with the external power transmission network.

[0228] In specific implementation, the active power of any micro-grid to the power transmission line within a specified time can be used as the transmission cost of any micro-grid, or can be determined according to other ways, which can be set according to actual application requirements.

[0229] After that, the active power loss within the specified time is determined based on the total power transmission line topology level, power transmission line resistance and power transmission line current.

[0230] Among them, the total power transmission line topology level describes the connection mode and structure of the power transmission line network, and different topology structures will affect the distribution and transmission efficiency of power. The power transmission line resistance is an important parameter of the power transmission line, which represents the resistance encountered by the current flowing in the power transmission line. The size of the line resistance is related to the material, length, cross-sectional area, etc. of the power transmission line. The power transmission line current is the driving force of energy transmission, and the size of the current determines the rate of energy transmission in the power transmission line.

[0231] The active power loss (P_loss) can usually be calculated by the following formula: P_loss = I 2 × R

[0232] Among them, I is the effective value of the power transmission line current (in amperes), and R is the power transmission line resistance (in ohms). This formula is based on Ohm's law and Joule's law, which describes the heat generated by the current flowing in the resistance, i.e. the active power loss.

[0233] The total power transmission line topology level can affect the distribution and flow of current in the power transmission line, thereby affecting the power loss. For example, in a star topology structure, the current may be more concentratedly flowing through certain key nodes and lines, resulting in an increase in power loss at these locations. Therefore, when designing and optimizing the micro-grid, the influence of the topology level on the power loss needs to be considered, and a suitable topology structure needs to be selected to reduce the power loss.

[0234] The latest current electricity price is the electricity price that meets the conditions that the power generation cost is less than a first preset constant, the active power loss is less than a second preset constant, and the transmission cost is minimized.

[0235] The first preset constant and the second preset constant can be set according to specific application requirements.

[0236] In the determination of the latest current electricity price, the disclosure comprehensively considers multiple factors such as power generation cost, active power loss, and transmission cost. Among them, the power generation cost needs to be lower than the first preset constant to ensure economy; the active power loss needs to be lower than the second preset constant to ensure energy efficiency; at the same time, the transmission cost needs to be minimized to ensure the efficient operation of the power grid. The comprehensive application of these conditions aims to realize the rationality, economy and sustainability of the electricity price setting.

[0237] Based on the above step S340, the disclosure obtains the latest current electricity price of any microgrid according to the power generation cost, transmission cost and active power loss of the microgrid. How to use the power generation cost, transmission cost and active power loss of any microgrid to determine is specific, and the specific way can be set according to actual application requirements.

[0238] In step S350, the latest current electricity price is sent to the energy regulation center; and the power consumption surplus or deficit amount in the future preset time is sent to the energy regulation center.

[0239] In step S360, a microgrid coordinated power consumption regulation scheme returned by the energy regulation center is received.

[0240] In step S370, power consumption regulation is performed according to the microgrid coordinated power consumption regulation scheme.

[0241] The microgrid coordinated power consumption regulation scheme can be a power allocation strategy, for example: power is transmitted from a microgrid with power surplus to a microgrid with power shortage, and at the same time, based on the consideration of electricity price factors, the cost of power allocation is optimized, that is, power is purchased when the electricity price is low, or unnecessary power consumption is reduced when the electricity price is high, or power is purchased from a microgrid with a lower electricity price, thereby balancing power supply and demand while optimizing the cost in the power shortage state.

[0242] FIG. 5 shows a method flowchart for optimizing a time attention mechanism-based LRCN double-layer network combination model according to an embodiment of the disclosure. As shown in FIG. 5, the following steps S510-S540 are included:

[0243] In step S510, the current actual electrical load and the current actual thermal load are obtained.

[0244] The current actual electric load and the current actual thermal load can be actual electric load and actual thermal load of a predicted day.

[0245] In step S520, an electric load deviation is obtained according to the current actual electric load and the electric load prediction value.

[0246] In step S530, a thermal load deviation is obtained according to the current actual thermal load and the thermal load prediction value.

[0247] If the current actual electric load and the current actual thermal load are actual electric load and actual thermal load of a predicted day, the electric load deviation is the difference between the average daily electric load prediction value and the current actual electric load; similarly, the thermal load deviation is the difference between the average daily thermal load prediction value and the current actual thermal load.

[0248] In step S540, when the electric load deviation and / or the thermal load deviation do not meet the preset requirements, the parameter migration of the LRCN double-layer network combination model based on the time attention mechanism is performed based on the online learning framework of the migration learning. Specifically, it includes:

[0249] If the difference between the average value of the historical electric load and the current actual electric load is greater than the preset electric load difference threshold, the first load feature extraction channel parameter is updated; if the difference between the average value of the historical thermal load and the current actual thermal load is greater than the preset thermal load difference threshold, the second load feature extraction channel parameter is updated; the current temperature and the historical temperature are obtained, and if the difference between the average value of the historical temperature and the current temperature is greater than the preset temperature difference threshold, the first load feature extraction channel parameter and the second load feature extraction channel parameter are updated.

[0250] The current temperature can be the temperature at the current time when the model is updated, and the historical data can be all available historical temperatures. The closer the historical data is to the current time, the higher the weight proportion is, and the greater the influence on the average value of the historical temperature is.

[0251] FIG. 6 shows another source network load storage collaborative intelligent power utilization method flow chart according to an embodiment of the present disclosure. The source network load storage collaborative intelligent power utilization method is applied to an energy regulation center, which is in communication connection with an energy regulation unit. The energy regulation unit is arranged in each microgrid in a microgrid cluster. The energy regulation units in each microgrid in the microgrid cluster are in communication connection. As shown in FIG. 6, the source network load storage collaborative intelligent power utilization method includes the following steps S610-S630:

[0252] In step S610, the power surplus / deficit amount in a future preset time sent by the energy regulation unit is received; and the latest current electricity price sent by the energy regulation unit is received.

[0253] In step S620, according to the power surplus / deficit amount of each microgrid in the microgrid cluster within the future preset time and the latest current electricity price, a microgrid coordinated power consumption regulation scheme is generated using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm. FIG. 7 shows a flowchart of a method for generating a microgrid coordinated power consumption regulation scheme based on a Pareto frontier curve and a fuzzy algorithm according to an embodiment of the present disclosure, specifically including the following steps S621-S625:

[0254] In step S621, the power surplus / deficit amount and the latest current electricity price are fuzzified using the fuzzy algorithm; the fuzzified power surplus / deficit amount and the latest current electricity price are converted into membership values in a fuzzy set through a predefined membership function.

[0255] In step S622, according to a preset parameter, a corresponding weight is assigned to each of a plurality of predefined optimization objectives; the optimization objectives include minimizing power consumption cost, maximizing energy utilization efficiency, or reducing environmental impact.

[0256] In step S623, according to the plurality of optimization objectives and the corresponding weights of each optimization objective, a multi-objective optimization algorithm is used to iteratively calculate the membership values in the fuzzy set, generating a plurality of preliminary power consumption regulation schemes; the multi-objective optimization algorithm includes a particle swarm optimization algorithm or a genetic algorithm.

[0257] In step S624, the objective function values of each preliminary power consumption regulation scheme are evaluated, and the positions of each preliminary power consumption regulation scheme are marked on the Pareto frontier curve.

[0258] In step S625, the microgrid coordinated power consumption regulation scheme is selected or generated through a preset fusion strategy in combination with the Pareto frontier curve and the membership values in the fuzzy set; the preset fusion strategy includes weighted fusion.

[0259] In step S630, the microgrid source-grid-load-storage coordinated power consumption intelligent regulation device sends the coordinated power consumption regulation scheme to the energy regulation unit in any microgrid.

[0260] FIG. 8 shows a structural block diagram of a source-grid-load-storage collaborative intelligent power utilization regulation device according to an embodiment of the present disclosure. The source-grid-load-storage collaborative intelligent power utilization regulation device is arranged in an energy regulation unit, which is arranged in each micro-grid in a micro-grid cluster, is in communication connection with an energy regulation center in the micro-grid cluster, and is in communication connection with the energy regulation units in each micro-grid in the micro-grid cluster.As shown in Figure 8, the intelligent control device in the energy regulation unit arranged in any micro-grid comprises: a parameter acquisition module arranged to acquire historical power consumption information of the any micro-grid and meteorological parameters in a future preset time; the historical power consumption information comprises historical electrical load and historical thermal load, and the meteorological parameters comprise one or more of the following parameters: illumination intensity, temperature, wind speed, wind direction, and relative humidity; a power consumption prediction module arranged to jointly predict power consumption of the any micro-grid in the future preset time using a LRCN double-layer network combination model based on a time attention mechanism according to the historical power consumption information and the meteorological parameters; the LRCN double-layer network combination model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module, and a multi-source data feature sharing layer module; the jointly predicting power consumption of the any micro-grid in the future preset time using the LRCN double-layer network combination model based on the time attention mechanism according to the historical power consumption information and the meteorological parameters comprises: processing the historical electrical load using a first load feature extraction channel to obtain feature representation of the historical electrical load; processing the historical thermal load using a second load feature extraction channel to obtain feature representation of the historical thermal load; processing temperature in the meteorological parameters, the feature representation of the historical electrical load, and the feature representation of the historical thermal load using the multi-source data feature sharing layer module to obtain electrical load prediction values and thermal load prediction values of the any micro-grid in the future preset time; obtaining power consumption of the any micro-grid in the future preset time according to the electrical load prediction values and the thermal load prediction values; a power consumption surplus or deficit calculation module arranged to determine power generation of the any micro-grid in the future preset time according to the meteorological parameters; determine power consumption surplus or deficit of the any micro-grid in the future preset time according to current energy storage capacity in the any micro-grid and the power generation in the future preset time and the power consumption in the future preset time; a latest current price calculation module arranged to acquire a latest current price of the any micro-grid according to power generation cost, transmission cost, and active power loss of the any micro-grid; a parameter sending module arranged to send the power consumption surplus or deficit in the future preset time to the energy regulation center; send the latest current price to the energy regulation center; a regulation scheme receiving module arranged to receive a micro-grid coordinated power consumption regulation scheme returned by the energy regulation center; a regulation scheme executing module arranged to perform power consumption regulation according to the micro-grid coordinated power consumption regulation scheme.

[0261] In the embodiments of the present disclosure, the parameter acquisition module, the power consumption prediction module, the power surplus / deficit calculation module, the latest current price calculation module, the parameter sending module, the regulation scheme receiving module and the regulation scheme executing module can be one or more processors, controllers or chips with a communication interface capable of implementing a communication protocol, and can also include a memory and related interfaces, a system transmission bus, etc. if necessary; the processor, controller or chip executes program-related codes to realize corresponding functions. Alternatively, the parameter acquisition module, the power consumption prediction module, the power surplus / deficit calculation module, the latest current price calculation module, the parameter sending module, the regulation scheme receiving module and the regulation scheme executing module share an integrated chip or share a processor, controller, memory, etc. The shared processor, controller or chip executes program-related codes to realize corresponding functions.

[0262] FIG. 9 shows a structural block diagram of another source-grid-load-storage coordinated intelligent power regulation device according to an embodiment of the present disclosure. The source-grid-load-storage coordinated intelligent power regulation device is arranged in an energy regulation center, which is in communication connection with energy regulation units arranged in each micro-grid in a micro-grid cluster, and the energy regulation units in each micro-grid in the micro-grid cluster are in communication connection. As shown in FIG. 9, the intelligent regulation device includes a parameter receiving module arranged to receive power surplus / deficit in a future preset time sent by the energy regulation units, receive the latest current price sent by the energy regulation units, a regulation scheme generating module arranged to generate a micro-grid coordinated power regulation scheme using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power surplus / deficit in the future preset time and the latest current price of each micro-grid in the micro-grid cluster, and a regulation scheme sending module arranged to send the micro-grid coordinated power regulation scheme to the energy regulation units in any micro-grid.

[0263] In the embodiments of the present disclosure, the parameter receiving module, the regulation scheme generating module and the regulation scheme sending module can be one or more processors, controllers or chips with a communication interface capable of implementing a communication protocol, and can also include a memory and related interfaces, a system transmission bus, etc. if necessary; the processor, controller or chip executes program-related codes to realize corresponding functions. Alternatively, the parameter receiving module, the regulation scheme generating module and the regulation scheme sending module share an integrated chip or share a processor, controller, memory, etc. The shared processor, controller or chip executes program-related codes to realize corresponding functions.

[0264] FIG. 10 shows a structural block diagram of an electronic device according to an embodiment of the present disclosure. As shown in FIG. 10, the electronic device includes a memory and a processor; wherein the memory is configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method according to any one of the above method embodiments.

[0265] The present disclosure also provides a computer-readable storage medium, which can be a computer-readable storage medium included in the electronic device or the computer system in the above embodiments; or can be a computer-readable storage medium that exists separately and is not assembled into the device. The computer-readable storage medium stores one or more programs, which are executed by one or more processors to implement the method described in the present disclosure.

[0266] The present disclosure also provides a computer program product, which includes a computer program that is executed by a processor to implement the method according to any one of the present disclosure.

[0267] According to the technical scheme provided by the embodiments of the present disclosure, a source network load storage collaborative intelligent regulation system is provided, which includes an energy regulation center and an energy regulation unit arranged in a micro-grid. Each energy regulation unit is responsible for collecting historical power consumption information of the micro-grid, meteorological parameters and other data in a future preset time, and using a LRCN double-layer network combination model based on time attention mechanism to predict power consumption. At the same time, each energy regulation unit is also responsible for determining the power generation capacity, energy storage capacity and power surplus / deficit in the future preset time of the micro-grid, and sending relevant information to the energy regulation center. The energy regulation center uses a fusion multi-objective algorithm based on the Pareto frontier curve and fuzzy algorithm to generate a micro-grid collaborative power regulation scheme according to the power surplus / deficit and the latest current electricity price of each micro-grid, and sends the regulation scheme to each energy regulation unit for execution, ensuring the balance of energy supply and demand of the micro-grid. Thus, efficient, intelligent and fine management of energy in the micro-grid cluster is realized, energy consumption and cost are reduced, and strong support is provided for the sustainable development of the micro-grid.

[0268] The above description is merely preferred embodiments of the present disclosure and a description of the principles of the technology employed. Those skilled in the art will understand that the scope of the application involved in the present disclosure is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or equivalent features without departing from the inventive concept. For example, the above features can be replaced with technical features disclosed in the present disclosure (but not limited to) having similar functions to form technical solutions.

Claims

1. A source network load storage collaborative intelligent regulation system, characterized in that, The intelligent regulation system comprises an energy regulation center and an energy regulation unit arranged in each micro-grid in a micro-grid cluster, the energy regulation center is in communication connection with the energy regulation units in the micro-grids respectively, and the energy regulation units in the micro-grids in the micro-grid cluster are in communication connection, wherein: The energy regulation unit in any micro-grid in the micro-grid cluster is configured to: acquire historical power consumption information and meteorological parameters in a future preset time of the micro-grid; the historical power consumption information comprises historical electrical load and historical thermal load, and the meteorological parameters comprise one or more of the following parameters: illumination intensity, temperature, wind speed, wind direction, and relative humidity; jointly predict power consumption of the micro-grid in the future preset time by using a LRCN double-layer network combination model based on a time attention mechanism according to the historical power consumption information and the meteorological parameters; the LRCN double-layer network combination model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module, and a multi-source data feature sharing layer module; the jointly predicting the power consumption of the micro-grid in the future preset time by using the LRCN double-layer network combination model based on the time attention mechanism according to the historical power consumption information and the meteorological parameters comprises: processing the historical electrical load by using a first load feature extraction channel to obtain feature representation of the historical electrical load; processing the historical thermal load by using a second load feature extraction channel to obtain feature representation of the historical thermal load; processing the temperature in the meteorological parameters, the feature representation of the historical electrical load, and the feature representation of the historical thermal load by using the multi-source data feature sharing layer module to obtain electrical load prediction values and thermal load prediction values of the micro-grid in the future preset time; and obtaining power consumption of the micro-grid in the future preset time according to the electrical load prediction values and the thermal load prediction values; determining power generation of the micro-grid in the future preset time according to the meteorological parameters; determining power surplus or deficiency of the micro-grid in the future preset time according to current energy storage capacity in the micro-grid, power generation in the future preset time, and power consumption in the future preset time; and sending the power surplus or deficiency in the future preset time to the energy regulation center; acquiring a latest current price of the micro-grid according to generation cost, transmission cost, and active power loss of the micro-grid; and sending the latest current price to the energy regulation center; the energy regulation center is configured to generate a micro-grid cooperative power consumption regulation scheme by using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power surplus or deficiency in the future preset time and the latest current price of each micro-grid in the micro-grid cluster; and send the micro-grid cooperative power consumption regulation scheme to the energy regulation unit in any micro-grid; the energy regulation unit is further configured to perform power consumption regulation according to the micro-grid cooperative power consumption regulation scheme.

2. The intelligent regulation system of claim 1, wherein, The any micro-grid further comprises: a power consumption data management system DMS configured to store, manage and maintain the historical electrical load and the historical thermal load, the obtaining of the historical power consumption information of the any micro-grid comprising: obtaining the historical power consumption information through interface communication with the power consumption data management system DMS.

3. The intelligent regulation system of claim 1, wherein, The first load feature extraction channel comprises: a first convolutional neural network CNN module, a first long short-term memory network LSTM module and a first efficient multi-scale attention module, and the processing of the historical electrical load using the first load feature extraction channel to obtain the feature representation of the historical electrical load comprises: extracting local features of the historical electrical load using the first convolutional neural network CNN module through convolution operation to obtain the feature map of the historical electrical load; processing the feature map of the historical electrical load using the first long short-term memory network LSTM module to capture long-term and short-term dependencies in the historical electrical load based on the characteristics of the loop connection to obtain the hidden state vector of the historical electrical load; processing the hidden state vector of the historical electrical load using the first efficient multi-scale attention module to obtain the feature representation of the historical electrical load after attention mechanism weighting by giving different weights to different features; The second load feature extraction channel comprises: a second convolutional neural network CNN module, a second long short-term memory network LSTM module and a second efficient multi-scale attention module, and the processing of the historical thermal load using the second load feature extraction channel to obtain the feature representation of the historical thermal load comprises: extracting local features of the historical thermal load using the second convolutional neural network CNN module through convolution operation to obtain the feature map of the historical thermal load; processing the feature map of the historical thermal load using the second long short-term memory network LSTM module to capture long-term and short-term dependencies in the historical thermal load based on the characteristics of the loop connection to obtain the hidden state vector of the historical thermal load; processing the hidden state vector of the historical thermal load using the second efficient multi-scale attention module to obtain the feature representation of the historical thermal load after attention mechanism weighting by giving different weights to different features.

4. The intelligent regulation system of claim 1, wherein, The energy regulation unit is further configured to: obtain the current actual electrical load and the current actual thermal load; obtain the electrical load deviation according to the current actual electrical load and the electrical load prediction value; obtain the thermal load deviation according to the current actual thermal load and the thermal load prediction value; when the electrical load deviation and / or the thermal load deviation do not meet the preset requirements, performing parameter migration on the LRCN double-layer network combination model based on time attention mechanism using an online learning framework based on transfer learning, comprising: if the difference between the average value of the historical electrical load and the current actual electrical load is greater than the preset electrical load difference threshold, updating the first load feature extraction channel parameters; updating the second load feature extraction channel parameter if a difference between an average value of the historical heat load and the current actual heat load is greater than a preset heat load difference threshold value; obtaining a current temperature and a historical temperature, and updating the first load feature extraction channel parameter and the second load feature extraction channel parameter if a difference between an average value of the historical temperature and the current temperature is greater than a preset temperature difference threshold value.

5. The intelligent regulation system of claim 1, wherein, The meteorological parameter further comprises: horizontal surface solar radiation per unit time; The power generation amount in the future preset time is calculated by the following formula, E=H×P×K×T; wherein E is the power generation amount in the future preset time, H is the horizontal surface solar radiation per unit time, P is the system installation capacity, K is the comprehensive efficiency coefficient, and T is the future preset time.

6. The intelligent regulation system of claim 1, wherein, The power surplus / deficit amount in the future preset time is calculated by the following formula, Q=A+B-C; wherein Q is the power surplus / deficit amount in the future preset time, A is the current energy storage capacity in the any micro-grid, B is the power generation amount in the future preset time, and C is the power consumption amount in the future preset time. When Q is positive, the power surplus / deficit amount in the future preset time indicates that the any micro-grid is in a power surplus state; when Q is negative, the power surplus / deficit amount in the future preset time indicates that the any micro-grid is in a power deficit state; and when Q is zero, the power surplus / deficit amount in the future preset time indicates that the any micro-grid is in a power balance state. The latest current price of the any micro-grid is obtained according to the power generation cost, the transmission cost and the active power loss of the any micro-grid, comprising:

7. The intelligent regulation system of claim 1, wherein, calculating the power generation cost of the any micro-grid based on the active power, the current price, the total discharge power of the any micro-grid and the demand side response power consumption of the any micro-grid in a specified time; determining the transmission cost based on the active power from the any micro-grid to the power transmission line in the specified time; determining the active power loss in the specified time based on the total power transmission line topology level, the power transmission line resistance and the power transmission line current; the latest current price is a price that satisfies the power generation cost being less than a first preset constant, the active power loss being less than a second preset constant and the transmission cost being minimum. The micro-grid cooperative power consumption regulation scheme is generated by using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power surplus / deficit amount in the future preset time and the latest current price of each micro-grid in the micro-grid cluster, comprising:

8. The intelligent regulation system of claim 1, wherein, the power surplus / deficit amount and the latest current price are subjected to fuzzy processing by using the fuzzy algorithm; and the power surplus / deficit amount and the latest current price after the fuzzy processing are converted into membership values in a fuzzy set by a pre-defined membership function. ​ According to the preset parameters, corresponding weights are assigned to a plurality of predefined optimization objectives; the optimization objectives include: minimizing electricity cost, maximizing energy utilization efficiency or reducing environmental impact; According to the plurality of optimization objectives and the weights corresponding to each optimization objective, a multi-objective optimization algorithm is used to iteratively calculate the membership values in the fuzzy set, generating a plurality of preliminary electricity regulation schemes; the multi-objective optimization algorithm includes: particle swarm optimization algorithm or genetic algorithm; Evaluate the objective function values of each preliminary electricity regulation scheme, and mark the positions of each preliminary electricity regulation scheme on the Pareto frontier curve; According to the Pareto frontier curve and the membership values in the fuzzy set, the micro-grid collaborative electricity regulation scheme is selected or generated by a preset fusion strategy; the preset fusion strategy includes: weighted fusion.

9. A source, network, load and storage collaborative intelligent power utilization regulation method, characterized in that, The intelligent regulation method is applied to an energy regulation unit, which is arranged in each micro-grid in a micro-grid cluster, and is in communication connection with an energy regulation center in the micro-grid cluster; the energy regulation units in each micro-grid in the micro-grid cluster are in communication connection; the intelligent regulation method includes: The energy regulation unit in any micro-grid in the micro-grid cluster obtains historical electricity information of the any micro-grid and meteorological parameters in a future preset time; the historical electricity information includes: historical electrical load and historical thermal load, and the meteorological parameters include one or more of the following parameters: illumination intensity, temperature, wind speed, wind direction, relative humidity; According to the historical electricity information and the meteorological parameters, the electricity consumption of the any micro-grid in the future preset time is jointly predicted using a LRCN double-layer network combination model based on a time attention mechanism; the LRCN double-layer network combination model based on the time attention mechanism includes: a convolutional neural network (CNN) module, a long short-term memory (LSTM) module, an efficient multi-scale attention module, and a multi-source data feature sharing layer module; the electricity consumption of the any micro-grid in the future preset time is jointly predicted using the LRCN double-layer network combination model based on the time attention mechanism according to the historical electricity information and the meteorological parameters, including: processing the historical electrical load using a first load feature extraction channel to obtain a feature representation of the historical electrical load; processing the historical thermal load using a second load feature extraction channel to obtain a feature representation of the historical thermal load; processing the temperature in the meteorological parameters, the feature representation of the historical electrical load, and the feature representation of the historical thermal load using the multi-source data feature sharing layer module to obtain electrical load prediction values and thermal load prediction values of the any micro-grid in the future preset time; obtaining the electricity consumption of the any micro-grid in the future preset time according to the electrical load prediction values and the thermal load prediction values; determining power generation of the any micro-grid in the future preset time according to the meteorological parameter; determining power consumption surplus or deficit of the any micro-grid in the future preset time according to current energy storage capacity in the any micro-grid, power generation in the future preset time and power consumption in the future preset time; sending the power consumption surplus or deficit in the future preset time to the energy regulation center; obtaining the latest current electricity price of the any micro-grid according to power generation cost, transmission cost and active power loss of the any micro-grid; sending the latest current electricity price to the energy regulation center; receiving the micro-grid cooperative power consumption regulation scheme returned by the energy regulation center; carrying out power consumption regulation according to the micro-grid cooperative power consumption regulation scheme.

10. The intelligent conditioning method of claim 9, wherein, The first load feature extraction channel includes a first convolutional neural network (CNN) module, a first long short-term memory (LSTM) module and a first efficient multi-scale attention module. The first load feature extraction channel is used to process the historical electrical load to obtain a feature representation of the historical electrical load, including: The first CNN module is used to extract local features of the historical electrical load through convolution operation to obtain a feature map of the historical electrical load. The first LSTM module is used to process the feature map of the historical electrical load to capture long-term and short-term dependencies in the historical electrical load based on the characteristics of a recurrent connection to obtain a hidden state vector of the historical electrical load. The first efficient multi-scale attention module is used to process the hidden state vector of the historical electrical load by assigning different weights to different features to obtain a feature representation of the historical electrical load weighted by an attention mechanism. The second load feature extraction channel includes a second CNN module, a second LSTM module and a second efficient multi-scale attention module. The second load feature extraction channel is used to process the historical thermal load to obtain a feature representation of the historical thermal load, including: The second CNN module is used to extract local features of the historical thermal load through convolution operation to obtain a feature map of the historical thermal load. The second LSTM module is used to process the feature map of the historical thermal load to capture long-term and short-term dependencies in the historical thermal load based on the characteristics of a recurrent connection to obtain a hidden state vector of the historical thermal load. The second efficient multi-scale attention module is used to process the hidden state vector of the historical thermal load by assigning different weights to different features to obtain a feature representation of the historical thermal load weighted by an attention mechanism.

11. The intelligent conditioning method of claim 9, wherein, The intelligent regulation method further includes: obtaining a current actual electrical load and a current actual thermal load; obtaining an electrical load deviation according to the current actual electrical load and the electrical load prediction value; obtaining a thermal load deviation according to the current actual thermal load and the thermal load prediction value; When the electrical load deviation and / or the thermal load deviation does not meet the preset requirement, a parameter migration is performed on the LRCN double-layer network combination model based on the time attention mechanism by using an online learning framework based on migration learning, including: If the difference between the average value of the historical electrical load and the current actual electrical load is greater than a preset electrical load difference threshold, the first load feature extraction channel parameter is updated; If the difference between the average value of the historical thermal load and the current actual thermal load is greater than a preset thermal load difference threshold, the second load feature extraction channel parameter is updated; The current temperature and the historical temperature are obtained, and if the difference between the average value of the historical temperature and the current temperature is greater than a preset temperature difference threshold, the first load feature extraction channel parameter and the second load feature extraction channel parameter are updated.

12. The intelligent conditioning method of claim 9, wherein, The meteorological parameter further includes: the horizontal plane solar radiation amount per unit time; The power generation amount in the future preset time is calculated by the following formula, E=H×P×K×T; Wherein, E is the power generation amount in the future preset time, H is the horizontal plane solar radiation amount per unit time, P is the system installation capacity, K is the comprehensive efficiency coefficient, and T is the future preset time.

13. The intelligent conditioning method of claim 9, wherein, The power consumption surplus or deficit of the any micro-grid in the future preset time is determined according to the current energy storage capacity in the any micro-grid, the power generation amount in the future preset time and the power consumption amount in the future preset time, including: The power consumption surplus or deficit in the future preset time is calculated by the following formula, Q=A+B-C; Wherein, Q is the power consumption surplus or deficit in the future preset time, A is the current energy storage capacity in the any micro-grid, B is the power generation amount in the future preset time, and C is the power consumption amount in the future preset time; When the Q is positive, the power consumption surplus or deficit in the future preset time is used to indicate that the any micro-grid is in a power surplus state; when the Q is negative, the power consumption surplus or deficit in the future preset time is used to indicate that the any micro-grid is in a power deficiency state, and when the Q is zero, the power consumption surplus or deficit in the future preset time is used to indicate that the any micro-grid is in a power balance state.

14. The intelligent conditioning method of claim 9, wherein, The latest current price of the any micro-grid is obtained according to the power generation cost, the transmission cost and the active power loss of the any micro-grid, including: The power generation cost of the any micro-grid is calculated based on the active power of the any micro-grid, the current price, the total discharge power of the any micro-grid and the demand side response power consumption of the any micro-grid in a specified time; The transmission cost is determined based on the active power from the any micro-grid to the power transmission line in the specified time; The active power loss in the specified time is determined based on the total power transmission line topology level, the power transmission line resistance and the power transmission line current; The latest current price is the price that meets the conditions that the power generation cost is less than a first preset constant, the active power loss is less than a second preset constant and the transmission cost is the minimum.

15. A source, network, load and storage collaborative intelligent power utilization regulation method, characterized in that, The application is applied to an energy regulation center, which is in communication connection with an energy regulation unit, the energy regulation unit is arranged in each micro-grid in a micro-grid cluster, the energy regulation units in each micro-grid in the micro-grid cluster are in communication connection, and the intelligent regulation method comprises the following steps: Receiving the power surplus / deficit amount in a future preset time sent by the energy regulation unit; Receiving the latest current electricity price sent by the energy regulation unit; According to the power surplus / deficit amount in the future preset time and the latest current electricity price of each micro-grid in the micro-grid cluster, a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm is used to generate a micro-grid cooperative power regulation scheme; The micro-grid cooperative power regulation scheme is sent to the energy regulation unit in any micro-grid.

16. The intelligent conditioning method of claim 15, wherein, According to the power surplus / deficit amount in the future preset time and the latest current electricity price of each micro-grid in the micro-grid cluster, a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm is used to generate a micro-grid cooperative power regulation scheme, which comprises the following steps: The fuzzy algorithm is used to fuzz the power surplus / deficit amount and the latest current electricity price; and the power surplus / deficit amount and the latest current electricity price after the fuzzing are converted into membership values in a fuzzy set through a pre-defined membership function; According to preset parameters, corresponding weights are allocated to a plurality of optimization objectives; the optimization objectives include minimizing power consumption cost, maximizing energy utilization efficiency or reducing environmental impact; According to the plurality of optimization objectives and the weights corresponding to each optimization objective, a multi-objective optimization algorithm is used to iteratively calculate the membership values in the fuzzy set, to generate a plurality of preliminary power regulation schemes, and the multi-objective optimization algorithm includes a particle swarm optimization algorithm or a genetic algorithm; The target function values of each preliminary power regulation scheme are evaluated, and the positions of each preliminary power regulation scheme are marked on the Pareto frontier curve; The Pareto frontier curve and the membership values in the fuzzy set are combined, and a preset fusion strategy is used to select or generate the micro-grid cooperative power regulation scheme; and the preset fusion strategy includes weighted fusion.

17. A source network load storage collaborative intelligent regulation and control device, characterized in that, The intelligent regulation device is arranged in an energy regulation unit, the energy regulation unit is arranged in each micro-grid in a micro-grid cluster, is in communication connection with an energy regulation center in the micro-grid cluster, and is in communication connection with the energy regulation units in each micro-grid in the micro-grid cluster, and the intelligent regulation device in the energy regulation unit in any micro-grid comprises the following steps: The parameter acquisition module is arranged to acquire historical power consumption information and meteorological parameters in a future preset time of the any micro-grid; the historical power consumption information includes historical electrical load and historical thermal load, and the meteorological parameters include one or more of the following parameters: light intensity, temperature, wind speed, wind direction and relative humidity; The power consumption prediction module is configured to jointly predict the power consumption of any micro-grid in the future preset time according to the historical power consumption information and the meteorological parameters by using a LRCN double-layer network combination model based on a time attention mechanism; the LRCN double-layer network combination model based on the time attention mechanism comprises a convolutional neural network (CNN) module, a long short-term memory (LSTM) network module, an efficient multi-scale attention module, and a multi-source data feature sharing layer module; the power consumption of any micro-grid in the future preset time is jointly predicted according to the historical power consumption information and the meteorological parameters by using the LRCN double-layer network combination model based on the time attention mechanism, which comprises: processing the historical electric load by using a first load feature extraction channel to obtain a feature representation of the historical electric load; processing the historical thermal load by using a second load feature extraction channel to obtain a feature representation of the historical thermal load; processing the temperature in the meteorological parameters, the feature representation of the historical electric load, and the feature representation of the historical thermal load by using the multi-source data feature sharing layer module to obtain an electric load prediction value and a thermal load prediction value of any micro-grid in the future preset time; and obtaining the power consumption of any micro-grid in the future preset time according to the electric load prediction value and the thermal load prediction value; The power consumption surplus or deficit calculation module is configured to determine the power generation of any micro-grid in the future preset time according to the meteorological parameters; and determine the power consumption surplus or deficit of any micro-grid in the future preset time according to the current energy storage capacity in any micro-grid, the power generation in the future preset time, and the power consumption in the future preset time; The latest current electricity price calculation module is configured to obtain the latest current electricity price of any micro-grid according to the power generation cost, transmission cost, and active power loss of any micro-grid; The parameter sending module is configured to send the power consumption surplus or deficit in the future preset time to the energy regulation and control center; and send the latest current electricity price to the energy regulation and control center; The regulation scheme receiving module is configured to receive a micro-grid coordinated power consumption regulation scheme returned by the energy regulation and control center; The regulation scheme executing module is configured to perform power consumption regulation according to the micro-grid coordinated power consumption regulation scheme.

18. A source network load storage collaborative intelligent regulation and control device, characterized in that, The energy regulation and control unit is arranged in each micro-grid in a micro-grid cluster, and the energy regulation and control units in each micro-grid in the micro-grid cluster are communicatively connected; the intelligent regulation and control device comprises: The parameter receiving module is configured to receive the power consumption surplus or deficit in the future preset time sent by the energy regulation and control unit; and receive the latest current electricity price sent by the energy regulation and control unit; The regulation scheme generating module is configured to generate a micro-grid coordinated power consumption regulation scheme by using a fusion multi-objective algorithm based on a Pareto frontier curve and a fuzzy algorithm according to the power consumption surplus or deficit in the future preset time and the latest current electricity price of each micro-grid in the micro-grid cluster. The regulation scheme sending module is configured to send the micro-grid cooperative power utilization regulation scheme to the energy regulation unit in any micro-grid.

19. An electronic device, comprising: The computer program product comprises a memory and a processor; wherein the memory is configured to store one or more computer instructions; and wherein the one or more computer instructions are executed by the processor to implement the intelligent regulation method according to any one of claims 9-16.

20. A computer readable storage medium having stored thereon computer instructions, wherein, The computer program product comprises a memory and a processor; wherein the memory is configured to store one or more computer instructions; and wherein the one or more computer instructions are executed by the processor to implement the intelligent regulation method according to any one of claims 9-16.

21. A computer program product comprising a computer program, characterized in that, The computer program product comprises a memory and a processor; wherein the memory is configured to store one or more computer instructions; and wherein the one or more computer instructions are executed by the processor to implement the intelligent regulation method according to any one of claims 9-16.

Images