An energy scheduling method and device based on multi-objective model solving, a terminal device, and a storage medium

Abstract

The application discloses an energy scheduling method and device based on a multi-target model solution, a terminal device and a storage medium. The causal neural network can be used to predict the output scheduling strategy at the next moment. In the training process of the causal neural network, the causal coefficients in the neural network are constantly updated based on the comparison result of the predicted result and the actual result, so that the causal neural network can learn and capture the causal relationship between the historical output scheduling strategy and the future output scheduling strategy, better adapt to the uncertainty and dynamics of the energy supply and demand system, and improve the reliability of the prediction. Therefore, when solving the multi-target scheduling optimization model, the application can perform multi-target solving based on each accurate output scheduling strategy at the next moment to obtain an accurate target output scheduling strategy that can minimize the operation cost, provide a reliable data basis for subsequent scheduling optimization, and avoid the occurrence of energy shortage or surplus.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy scheduling, and in particular to an energy scheduling method and device based on multi-objective model solving, a terminal device and a storage medium. BACKGROUND

[0002] With the continuous coupling between natural gas, heat-cold networks and electrical systems, in the planning and implementation of the scheduling of each energy system, the utilization efficiency, economy, environmental protection and stability of each energy need to be considered comprehensively, and the problem of multiple indicators can be converted into a multi-objective problem. In a comprehensive energy system, both economy and carbon emissions need to be considered, and the energy conversion cost is also optimized, so there may be a mutual restriction or conflict between multiple objectives. Therefore, how to find a solution that can simultaneously satisfy or balance these objectives is a problem that needs to be solved.

[0003] The prior art often solves multi-objective problems by mathematical analysis and intelligent optimization algorithms, that is, the optimal scheduling strategy can be solved among multiple scheduling strategies. However, due to the significant intermittency and volatility of new energy such as solar energy and wind energy, and the randomness of user behavior, the prior art needs to overcome uncertainty when predicting future energy supply and demand through neural networks or models to generate multiple scheduling strategies. However, the existing model cannot fully capture the uncertainty and dynamics of the energy supply and demand system, and it is often difficult to accurately predict future energy scheduling strategies. Therefore, when solving the optimal scheduling strategy based on multiple predicted scheduling strategies, the optimal scheduling strategy obtained by solving multiple objectives based on inaccurate scheduling strategies may have deviations. This deviation will be amplified in subsequent energy scheduling and output control, resulting in energy shortages or surpluses. SUMMARY

[0004] The embodiments of the present application provide an energy scheduling method and device based on multi-objective model solving, a terminal device and a storage medium. By introducing a causal neural network, the energy supply and demand system can better adapt to the uncertainty and dynamics of the energy supply and demand system, improve the prediction stability and reliability of the optimal scheduling strategy, and effectively solve the problem of energy shortages or surpluses caused by deviations in the optimal scheduling strategy in the prior art.

[0005] An embodiment of the present application provides an energy scheduling method based on multi-objective model solving, comprising:

[0006] According to first basic data of the energy conversion system, second basic data of the combined cooling heating and power system and third basic data of the electricity-to-gas system, a scheduling optimization model for minimizing operation cost is constructed; wherein the first basic data comprises first output data and first operation cost data of the energy conversion system; the second basic data comprises second output data and second operation cost data of the combined cooling heating and power system; and the third basic data comprises third output data and third operation cost data of the electricity-to-gas system;

[0007] The historical output scheduling strategies corresponding to the current time are input into a preset causal neural network, so that the causal neural network predicts the output scheduling strategy of the next time corresponding to each current time; wherein the historical output scheduling strategy comprises historical output scheduling values corresponding to the energy conversion system, the combined cooling heating and power system and the electricity-to-gas system respectively;

[0008] Under the preset output constraint of the scheduling optimization model, the scheduling optimization model is solved based on the output scheduling strategies of the next time, to generate a target output scheduling strategy at the minimum operation cost;

[0009] According to the target output scheduling strategy, the output of the energy conversion system, the combined cooling heating and power system and the electricity-to-gas system is controlled respectively;

[0010] The generation of the causal neural network comprises:

[0011] Each sample output scheduling strategy corresponding to a sample time is input into the neural network as an input, and the actual output scheduling strategy of the next time corresponding to each sample time is output as an output, and the neural network is iteratively trained until the neural network converges, and the trained neural network is taken as the causal neural network;

[0012] At each training time, each sample output scheduling strategy corresponding to a sample time is input into the neural network, so that the neural network outputs the output scheduling strategy prediction result corresponding to each sample time; the output scheduling strategy prediction result corresponding to each sample time is compared with the actual output scheduling strategy of the next time, and the causal coefficient in the neural network is adjusted according to the comparison result;

[0013] The causal coefficient is used to represent the causal relationship between the sample output scheduling strategy corresponding to a sample time and the actual output scheduling strategy of the next time corresponding thereto.

[0014] Preferably, the energy conversion system is a gas turbine for realizing the conversion among power supply, heat supply and gas supply; the combined cooling heating and power system is a refrigeration machine device capable of realizing the integration of refrigeration, heat supply and power generation; and the electricity-to-gas system is an electricity-to-gas device for converting wind power into natural gas.

[0015] The first operation cost data includes power generation cost data of the gas turbine, carbon emission cost data of the gas turbine, and energy conversion cost data of the gas turbine; the second operation cost data includes conversion cost data of the refrigeration machine device, carbon emission cost data of the refrigeration machine device, and energy conversion cost data of the refrigeration machine device; and the third operation cost data includes conversion cost data of the electric-gas device, carbon emission cost data of the electric-gas device, and energy conversion cost data of the electric-gas device.

[0016] The scheduling optimization model for minimizing the operation cost is constructed according to the following formula:

[0017] F = argmin (F1, F2, F3) ;

[0018]

[0019]

[0020] Wherein, F is used to represent the scheduling optimization model for minimizing the total operation cost of the gas turbine, the refrigeration machine device, and the electric-gas device, F1 is used to represent the total start-up cost of the gas turbine, the refrigeration machine device, and the electric-gas device, F2 is used to represent the total carbon emission cost of the gas turbine, the refrigeration machine device, and the electric-gas device, and F3 is used to represent the total energy conversion cost of the gas turbine, the refrigeration machine device, and the electric-gas device. is the power generation cost of the gas turbine; is the conversion cost of the refrigeration machine device; is the energy generation cost of the refrigeration device; is the conversion cost of the electric-gas device; is the carbon emission cost of the gas turbine; is the carbon emission cost of the refrigeration machine device; is the carbon emission cost of the electric-gas device; is the energy conversion cost of the gas turbine operation; is the energy conversion cost of the combined cooling, heating and power system; is the energy conversion cost of the electric-gas device;

[0021] a, b and c are different efficiency coefficients of the micro gas turbine; P mt is the start-stop state of the gas turbine; H ac is the input heat power of the refrigeration machine device; COP ac is the refrigeration coefficient of the refrigeration machine device; D ac is the conversion coefficient of the refrigeration machine through electric energy; P ac is the operation energy consumption of the refrigeration machine device; η p2g is the conversion efficiency of the electric-gas device; Pp2g P is the power of the electric-gas conversion technology; a, b, g are the quadratic factor, linear factor and constant factor of the carbon emission of the gas turbine, respectively; r ac is the carbon emission factor of the refrigeration equipment; is the energy conversion factor of the refrigeration equipment; is the energy conversion factor of the electric-gas conversion equipment.

[0022] Preferably, the preset output constraint of the scheduling optimization model comprises:

[0023]

[0024] wherein, is a preset upper limit value of the output power of the gas turbine, is a preset upper limit value of the input thermal power of the refrigeration equipment, is a preset upper limit value of the power of the electric-gas conversion equipment.

[0025] Preferably, the neural network comprises a causal selection network and a causal prediction network.

[0026] The inputting of the sample output scheduling strategy corresponding to each sample moment into the neural network, so that the neural network outputs the output scheduling strategy prediction result corresponding to each sample moment, comprises:

[0027] The inputting of the sample output scheduling strategy corresponding to each sample moment into the neural network, so that the causal selection network extracts the strategy feature corresponding to each sample output scheduling strategy;

[0028] The causal prediction network predicts the output scheduling strategy prediction result corresponding to each sample moment according to the strategy feature.

[0029] Preferably, the comparing of the output scheduling strategy prediction result corresponding to each sample moment with the actual output scheduling strategy of the next moment, and the adjusting of the causal coefficient in the neural network according to the comparison result, comprises:

[0030] The comparing of the output scheduling strategy prediction result corresponding to each sample moment with the actual output scheduling strategy of the next moment, and the updating of the loss function corresponding to the neural network according to the comparison result; wherein the loss function is:

[0031]

[0032] wherein, L loss represents a loss value, S' is the output scheduling strategy prediction result corresponding to the sample moment; S is the actual output scheduling strategy of the next moment; m is the data size of the small batch sampling, and i is the sample index of the small batch sampling;

[0033] According to the updated loss function, a gradient value of each output scheduling strategy prediction result corresponding to a moment of each sample is calculated, wherein the gradient value is used to represent an adjustment direction and an adjustment size corresponding to a parameter in the neural network.

[0034] According to the gradient value corresponding to each output scheduling strategy prediction result, a causal coefficient in the neural network is adjusted.

[0035] Preferably, the calculation of the gradient value of each output scheduling strategy prediction result corresponding to the moment of each sample according to the updated loss function comprises:

[0036] According to the loss function, a loss value of each output scheduling strategy prediction result is calculated.

[0037] The loss value of each output scheduling strategy prediction result is multiplied by an activation function of the causal selection network and the causal prediction network respectively, so as to obtain a gradient value corresponding to each output scheduling strategy prediction result in the causal selection network and the causal prediction network respectively.

[0038] Preferably, the adjustment of the causal coefficient in the neural network according to the gradient value corresponding to each output scheduling strategy prediction result comprises:

[0039] According to the gradient value of each output scheduling strategy prediction result in the causal selection network and the causal prediction network respectively, the causal coefficients corresponding to the causal selection network and the causal prediction network are adjusted respectively.

[0040] On the basis of the method embodiments described above, the present application provides device embodiments corresponding thereto.

[0041] An embodiment of the present application provides an energy scheduling device based on a multi-target model solution, comprising a model construction module, a scheduling strategy prediction module, a multi-target solution module and an output control module.

[0042] The model construction module is used to construct a scheduling optimization model with the minimum operation cost according to first basic data of an energy conversion system, second basic data of a combined cooling heating and power system and third basic data of an electric-gas conversion system, wherein the first basic data comprises first output data and first operation cost data of the energy conversion system, the second basic data comprises second output data and second operation cost data of the combined cooling heating and power system, and the third basic data comprises third output data and third operation cost data of the electric-gas conversion system.

[0043] The scheduling strategy prediction module is configured to input historical output scheduling strategies corresponding to a plurality of current time points into a preset causal neural network, so that the causal neural network predicts an output scheduling strategy of a next time point corresponding to each current time point; wherein the historical output scheduling strategies include historical output scheduling values corresponding to an energy conversion system, a combined cooling heating and power system, and an electric-to-gas system respectively.

[0044] The multi-objective solving module is configured to solve the scheduling optimization model based on the output scheduling strategies of the plurality of next time points under preset output constraints of the scheduling optimization model, to generate a target output scheduling strategy at a minimum running cost.

[0045] The output control module is configured to control outputs of the energy conversion system, the combined cooling heating and power system, and the electric-to-gas system respectively according to the target output scheduling strategy.

[0046] The generation of the causal neural network includes:

[0047] Each sample output scheduling strategy corresponding to a sample time point is taken as an input of a neural network, an actual output scheduling strategy of a next time point corresponding to each sample time point is taken as an output of the neural network, the neural network is iteratively trained until the neural network converges, and the trained neural network is taken as the causal neural network.

[0048] During each training, each sample output scheduling strategy corresponding to a sample time point is input into the neural network, so that the neural network outputs an output scheduling strategy prediction result corresponding to each sample time point; the output scheduling strategy prediction result corresponding to each sample time point is compared with an actual output scheduling strategy of a next time point corresponding to the sample time point, and a causal coefficient in the neural network is adjusted according to a comparison result.

[0049] The causal coefficient is used to represent a causal relationship between the sample output scheduling strategy corresponding to the sample time point and the actual output scheduling strategy of the next time point corresponding to the sample time point.

[0050] On the basis of the method embodiment, the application correspondingly provides a terminal device embodiment.

[0051] Another embodiment of the application provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the energy scheduling method based on multi-objective model solving in the above-mentioned embodiment of the application when executing the computer program.

[0052] On the basis of the method embodiment, the application correspondingly provides a storage medium embodiment.

[0053] Another embodiment of the present application provides a storage medium, the computer readable storage medium comprises a stored computer program, wherein the computer readable storage medium controls the device where the computer readable storage medium is located to execute the energy scheduling method based on multi-target model solving of the above-mentioned embodiment of the present application when the computer program runs.

[0054] By implementing the present application, the following beneficial effects are achieved:

[0055] The embodiment of the present application provides an energy scheduling method and device based on multi-target model solving, terminal equipment and storage medium. The energy scheduling method of the present application can predict the output scheduling strategy at the next moment by using the causal neural network, and in the training process of the causal neural network, the causal coefficients in the neural network are constantly updated based on the comparison result of the predicted result and the actual result, so that the causal neural network can learn and capture the causal relationship between the historical output scheduling strategy and the future output scheduling strategy. Therefore, when applied, the future (i.e. the next moment) output scheduling strategy can be more accurately predicted. Therefore, by introducing the causal neural network, the present application can better adapt to the uncertainty and dynamics of the energy supply and demand system, and improve the stability and reliability of the prediction. Compared with the prior art, when solving the multi-target scheduling optimization model, the present application can perform multi-target solving based on the accurate output scheduling strategy at the next moment to obtain an accurate target output scheduling strategy that can minimize the operation cost, and provides a reliable data basis for subsequent scheduling optimization, so that the output can be adjusted based on the target output scheduling strategy with small deviation from the actual situation, thereby avoiding the occurrence of energy shortage or excess. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a flowchart of an energy scheduling method based on multi-target model solving provided by an embodiment of the present application.

[0057] Figure 2 is a schematic diagram of a causal neural network structure provided by an embodiment of the present application.

[0058] Figure 3 is a schematic diagram of parameter self-updating based on meta-learning provided by an embodiment of the present application.

[0059] Figure 4 is a structural schematic diagram of an energy scheduling device based on multi-target model solving provided by an embodiment of the present application. DETAILED DESCRIPTION

[0060] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0061] As shown in Figure 1 Fig. 1 is a flow diagram of an energy scheduling method based on multi-objective model solving provided by an embodiment of the present application, and the energy scheduling method based on multi-objective model solving comprises the following steps:

[0062] Step S1: constructing a scheduling optimization model for minimizing operation cost according to first basic data of an energy conversion system, second basic data of a combined cooling heating and power system and third basic data of an electric-gas conversion system; wherein the first basic data comprises first output data and first operation cost data of the energy conversion system; the second basic data comprises second output data and second operation cost data of the combined cooling heating and power system; and the third basic data comprises third output data and third operation cost data of the electric-gas conversion system;

[0063] Step S2: inputting a plurality of historical output scheduling strategies corresponding to current time into a preset causal neural network, so that the causal neural network predicts an output scheduling strategy of a next time corresponding to each current time; wherein the historical output scheduling strategy comprises historical output scheduling values corresponding to the energy conversion system, the combined cooling heating and power system and the electric-gas conversion system respectively;

[0064] Wherein, the generation of the causal neural network comprises:

[0065] Taking the sample output scheduling strategy corresponding to each sample time as the input of the neural network, taking the actual output scheduling strategy of the next time corresponding to each sample time as the output of the neural network, iteratively training the neural network, and taking the trained neural network as the causal neural network when the neural network converges;

[0066] In each training, inputting the sample output scheduling strategy corresponding to each sample time into the neural network, so that the neural network outputs the output scheduling strategy prediction result corresponding to each sample time; comparing the output scheduling strategy prediction result corresponding to each sample time with the actual output scheduling strategy of the next time, and adjusting the causal coefficient in the neural network according to the comparison result;

[0067] Wherein, the causal coefficient is used to represent the causal relationship between the sample output scheduling strategy corresponding to the sample time and the actual output scheduling strategy of the next time corresponding thereto;

[0068] Step S3: based on the output scheduling strategy of the next time, the scheduling optimization model is solved under the preset output constraint of the scheduling optimization model, to generate the target output scheduling strategy at the minimum running cost;

[0069] Step S4: according to the target output scheduling strategy, the output of the energy conversion system, the combined cooling heating and power system and the electric-to-gas system is controlled respectively.

[0070] For step S1, in a preferred embodiment, the energy conversion system can be a gas turbine for realizing the conversion among power supply, heat supply and gas supply; the combined cooling heating and power system can be a refrigeration machine device for realizing the integration of refrigeration, heat supply and power generation; and the electric-to-gas system can be an electric-to-gas device for converting wind power into natural gas.

[0071] Further, the first running cost data mainly includes the power generation cost data of the gas turbine, the carbon emission cost data of the gas turbine and the energy conversion cost data of the gas turbine; the second running cost data mainly includes the conversion cost data of the refrigeration machine device, the carbon emission cost data of the refrigeration machine device and the energy conversion cost data of the refrigeration machine device; and the third running cost data mainly includes the conversion cost data of the electric-to-gas device into natural gas, the carbon emission cost data of the electric-to-gas device and the energy conversion cost data of the electric-to-gas device.

[0072] The micro gas turbine is the main device for realizing the conversion among power supply, heat supply and gas supply, and its mathematical model can be expressed as follows:

[0073]

[0074] wherein, is the power generation cost of the gas turbine; P mt is the start-stop state of the gas turbine; a, b and c are different efficiency coefficients of the micro gas turbine.

[0075] The combined cooling heating and power system is the main device for realizing the integration of refrigeration, heat supply and power generation, and the mathematical model of the absorption refrigeration machine device is shown as follows:

[0076]

[0077] wherein, is the conversion cost of the refrigeration machine device; H ac is the input heat power of the refrigeration machine device; COP ac is the refrigeration coefficient of the refrigeration machine device, is the energy generation cost of the refrigeration device; P mt is the start-stop state of the gas turbine, D ac is the conversion coefficient of the refrigeration machine through electric energy, η is the conversion cost of the electric-to-gas equipment for converting into natural gas; η p2 g is the conversion efficiency of the electric-to-gas equipment; α, β, γ are quadratic factor, linear factor and constant factor of carbon emission of the gas turbine respectively; P ac η is the operation energy consumption of the refrigeration equipment; r ac η is the carbon emission factor of the refrigeration equipment.

[0078] The electric-to-gas technology is a technology of converting wind power into natural gas, and the mathematical model of the corresponding electric-to-gas equipment is as follows:

[0079]

[0080] η is the operation energy consumption of the refrigeration equipment; r η is the energy conversion cost of the combined cooling heating and power system; η is the energy conversion cost of the electric-to-gas equipment; η is the energy conversion factor of the refrigeration equipment; η is the energy conversion factor of the electric-to-gas equipment. Illustratively, the energy conversion cost of the combined cooling heating and power system is the carbon emission cost data of the electric-to-gas equipment in the third operation cost data mentioned above.

[0081] Through the above-mentioned electric, gas, cold (hot) gas system mathematical model, the scheduling optimization model with the targets of economy, carbon emission reduction and energy conversion cost is as follows:

[0082] F = argmin (F1, F2, F3);

[0083]

[0084] F is used to represent the scheduling optimization model for minimizing the total operation cost of the gas turbine, the refrigeration equipment and the electric-to-gas equipment, F1 is used to represent the total start-up cost of the gas turbine, the refrigeration equipment and the electric-to-gas equipment, F2 is used to represent the total carbon emission cost of the gas turbine, the refrigeration equipment and the electric-to-gas equipment, and F3 is used to represent the total energy conversion cost of the gas turbine, the refrigeration equipment and the electric-to-gas equipment; η is the power generation cost of the gas turbine; η is the conversion cost of the refrigeration equipment; η is the energy generation cost of the refrigeration equipment; η is the conversion cost of the electric-to-gas equipment for converting into natural gas; η is the carbon emission cost of the gas turbine; η is the carbon emission cost of the refrigeration equipment; η is the carbon emission cost of the electric-to-gas equipment; η is the energy conversion cost of the gas turbine operation; η is the energy conversion cost of the combined cooling heating and power system; The energy conversion cost of the electricity-to-gas device.

[0085] Further, the preset output constraint of the scheduling optimization model is a processing constraint corresponding to the energy conversion system, the combined heat and power system, and the electricity-to-gas system, respectively, and thus,

[0086]

[0087] wherein, the preset upper limit value of the output power of the gas turbine, the preset upper limit value of the input heat power of the refrigeration device, the preset upper limit value of the power of the electricity-to-gas device.

[0088] For step S2, in a preferred embodiment, although the existing technologies such as mathematical analysis method and intelligent optimization algorithm have achieved certain results in solving multi-objective problems, the accuracy of the prediction and scheduling strategy is challenged when facing the intermittency and volatility of new energy and the randomness of user behavior. Moreover, due to these characteristics of new energy and the randomness of user behavior, the uncertainty and dynamics of the energy supply and demand system are increased, making it difficult for the existing model to fully capture these changes, thereby affecting the accuracy of prediction and scheduling.

[0089] When multi-objective solving is based on inaccurate scheduling strategies, the optimal scheduling strategy obtained is likely to be biased, so that in the subsequent energy scheduling and output control, due to the biased output supply, energy shortage or surplus problems will occur. Energy shortage may lead to insufficient power supply, heat supply or cooling supply in some areas, affecting the life of residents and industrial production; and energy surplus will cause resource waste and environmental pollution.

[0090] In order to overcome these problems, before solving the above multi-objective model (i.e., the scheduling optimization model), the present application can predict accurate scheduling strategies through the preset causal neural network, and then solve the multi-objective model based on these predicted scheduling strategies, to ensure that the subsequent solved target output scheduling strategy is more accurate and reduces the deviation from the actual output scheduling.

[0091] In the embodiment of the present application, a number of historical output scheduling strategies corresponding to the current time can be input into the preset causal neural network, so that the causal neural network predicts the output scheduling strategy of the next time corresponding to each current time;

[0092] The present application provides a real-time scheduling strategy distribution prediction scheme for comprehensive energy based on causal deep learning, such as Figure 2The structural diagram of the illustrated causal neural network, the neural network (i.e., the causal neural network) comprises a causal selection network and a causal prediction network; wherein the causal selection network and the causal prediction network are both composed of multiple hidden layers.

[0093] In order to improve the prediction accuracy of the causal neural network of the present application, during the training process of the causal neural network, when the sample output scheduling strategy corresponding to each sample time is input into the neural network to make the neural network output the output scheduling strategy prediction result corresponding to each sample time, specifically includes:

[0094] The sample output scheduling strategy corresponding to each sample time is input into the neural network to make the causal selection network extract the strategy feature corresponding to each sample output scheduling strategy;

[0095] The causal prediction network predicts the output scheduling strategy prediction result corresponding to each sample time according to the strategy feature.

[0096] Specifically, the causal selection network can extract historical time sequence features, assuming that the prediction time is T = (t1, t2,..., tn), and the sample historical output scheduling strategy of the energy supplier can be represented as:

[0097] {S = {(F1, t1), (F2, t2),..., (Fn, tn)}}; n n}};

[0098] The output scheduling strategy prediction result corresponding to each sample time predicted by the causal prediction network can be represented as:

[0099] {S' = {(F1', t1'), (F2', t'2),..., (Fn', t'n)}}; n n}}.

[0100] After obtaining the output scheduling strategy prediction result corresponding to each sample time, the output scheduling strategy prediction result corresponding to each sample time can be compared with the actual output scheduling strategy of the next time;

[0101] First, update the loss function corresponding to the neural network according to the comparison result; wherein the loss function is:

[0102]

[0103] Wherein, L loss represents the loss value, S' is the output scheduling strategy prediction result corresponding to the sample time; S is the actual output scheduling strategy of the next time; m is the data size of the small batch sampling, and i is the sample index of the small batch sampling; ​​

[0104] Then, based on the updated loss function, backpropagation is performed, and the parameters of the network are updated using the ReLU function to accurately predict the output scheduling strategy at the next time point, specifically:

[0105] According to the loss function, the loss value of each output scheduling strategy prediction result is calculated.

[0106] The loss value of each output scheduling strategy prediction result is multiplied by the activation function of the causal selection network and the causal prediction network, respectively, to obtain the gradient value corresponding to each output scheduling strategy prediction result in the causal selection network and the causal prediction network, respectively.

[0107] The gradient value is used to represent the adjustment direction and adjustment size of the parameters in the neural network. The gradient value reflects the influence of the difference between the prediction result and the actual result on each parameter in the network. In the backpropagation process, these gradient values are used to guide the update of the network parameters. For the causal selection network and the causal prediction network of the present application, their respective gradient values reflect the contribution degree of the final prediction result and the adjustment direction.

[0108] Further, the causal coefficients corresponding to the causal selection network and the causal prediction network can be adjusted according to the gradient value of each output scheduling strategy prediction result in the causal selection network and the causal prediction network, respectively. Illustratively, the parameters in the network, including the causal coefficients and other network weights, can be updated according to the calculated gradient, and the updated direction is the opposite direction of the gradient, because the goal is to minimize the loss function. Through multiple iterations of training, the network parameters are constantly optimized until the preset number of training rounds is reached or other stopping conditions are met.

[0109] It can be understood that the causal neural network can capture the causal relationship in the data. By updating the parameters of the network (including but not limited to the causal coefficients), the causal neural network can better learn and understand the causal relationship between the input data (sample output scheduling strategy) and the output data (output scheduling strategy prediction result), so that the network can be closer to the actual result in subsequent prediction. Illustratively, compared with traditional Sigmoid or Tanh activation functions, the ReLU function has a constant gradient (i.e. 1) when the input is positive, which helps to alleviate the common gradient vanishing problem in deep neural networks.

[0110] In the causal neural network, the causal selection network is responsible for extracting historical time series features, and the causal prediction network makes predictions based on these features. By adjusting the causal coefficients, the influence of the two parts on the prediction result can be balanced to ensure that they can fully play their respective roles.

[0111] When adjusting the causal coefficients, the application iteratively updates various parameters in the network through an optimization algorithm (such as a gradient descent algorithm). Specifically, according to the gradient value and the learning rate of the network (a hyperparameter that controls the step size of parameter updates), the update amount of the causal coefficients can be calculated and applied to the network. This process is repeated until the network converges or other stopping conditions are met.

[0112] By adjusting the causal coefficients, the network can better adapt to specific data distributions and task requirements, thereby improving its performance in practical applications. The application updates the causal coefficients in the neural network based on the comparison between the predicted results and the actual results during the training process of the causal neural network, so that the causal neural network can learn and capture the causal relationship between historical output scheduling strategies and future output scheduling strategies. Therefore, the application can more accurately predict the future (i.e., the next time) output scheduling strategy when applied. Therefore, by introducing a causal neural network, the application can better adapt to the uncertainty and dynamics of the energy supply and demand system, improving the stability and reliability of the prediction.

[0113] In a preferred embodiment, to improve the accuracy of the predicted scheduling strategy and speed up the training of the neural network, the application introduces the idea of meta-learning to compare the similarity of historical scheduling strategies and real-time predicted scheduling strategies, to better optimize the predicted output scheduling strategy at the next time. As shown in the parameter self-updating schematic diagram based on meta-learning Figure 3 , in meta-learning, the historical time series features of energy suppliers extracted by the causal selection network are first pre-trained;

[0114] The historical time series features can be: D tr = {(F1, t1), (F2, t2),..., (F n , t n )};

[0115] The training set is set to D tr = {(F1, t1), (F2, t2),..., (F n , t n )}; then the causal prediction network based on the test set takes the test set as input, and finally obtains the probability value between the test set and the training value as output. The greater the probability, the more similar the test set and the training value; and the test set updates the parameters φ in the network in an adaptive manner until the network converges. Thus, the application improves the extraction ability of the causal selection network and the prediction ability of the causal prediction network through adaptive updating of the parameters, and enables the causal neural network to accurately predict the output scheduling strategy of the energy supply even with few samples.

[0116] For steps S3 and S4, in a preferred embodiment, when solving the multi-objective scheduling optimization model, the present application can perform multi-objective solving based on the accurate next-time output scheduling strategy of each target, and when solving the scheduling optimization model, an accurate target output scheduling strategy that can minimize the operation cost can be generated, providing a reliable data basis for subsequent scheduling optimization.

[0117] Thus, the output of the integrated multi-energy supply system can be adjusted based on the target output scheduling strategy with less deviation from the actual situation, thereby avoiding energy shortage or surplus.

[0118] Specifically, by adjusting the power (i.e., output) of the equipment, fine management of energy use can be achieved, thereby further optimizing the energy use strategy. For example, during peak electricity consumption periods, the power of non-critical equipment can be reduced to alleviate the load on the power grid and improve the stability of the power grid; during off-peak electricity consumption periods, the power of the equipment can be increased to fully utilize the remaining capacity of the power grid. The present application can predict the output scheduling strategy at the next time according to the historical output scheduling strategy at the current time, which enables the energy scheduling system to respond in real time to changes and quickly adjust the scheduling strategy to adapt to changing energy demands, which can help to achieve peak-shaving use of energy and improve energy utilization efficiency.

[0119] In a preferred embodiment, in the integrated multi-energy supply system, the scheduling strategy can also contain the plans and decisions of each energy supplier to produce, convert, distribute, and use energy according to factors such as energy demand, supply conditions, equipment status, and market conditions. Thus, the scheduling strategy aims to optimize the operation of the energy system to meet energy demand while achieving economic, environmental, and stability goals. When the historical scheduling strategy of the energy supplier is associated with the output data, it generally means that the actual effect of the historical scheduling strategy and the corresponding output data are considered during the formulation and execution of the scheduling strategy. In embodiments of the present application, the output data refers to the amount of energy or power output of energy supply equipment (such as gas turbines, wind turbines, solar photovoltaic panels, etc.) under a specific scheduling strategy.

[0120] The present application realizes the collaborative optimization and scheduling of multiple energy systems by combining scheduling optimization models and causal neural network prediction. Not only can the operating characteristics and costs of each system be considered, but also historical data can be used to predict future output scheduling strategies, thereby improving energy utilization efficiency and reducing operating costs.

[0121] In summary, the present application realizes the real-time scheduling strategy distribution prediction and multi-objective solving method of comprehensive energy by causal deep learning, solves the problems that the prior art cannot cope with complex systems and cannot dynamically maintain multiple target optimization, improves the prediction accuracy, optimizes the scheduling strategy, and enables the comprehensive energy system to ensure safe and stable operation while minimizing multiple targets such as economy, carbon emissions, and energy conversion cost.

[0122] As shown in the above various embodiments of the energy scheduling method based on multi-objective model solving, the present application correspondingly provides device embodiments; Figure 3

[0123] An embodiment of the present application provides an energy scheduling device based on multi-objective model solving, comprising a model construction module, a scheduling strategy prediction module, a multi-objective solving module, and an output control module.

[0124] The model construction module is configured to construct a scheduling optimization model for minimizing operation cost according to first basic data of an energy conversion system, second basic data of a combined heat and power system, and third basic data of an electric-to-gas system; wherein the first basic data comprises first output data and first operation cost data of the energy conversion system; the second basic data comprises second output data and second operation cost data of the combined heat and power system; and the third basic data comprises third output data and third operation cost data of the electric-to-gas system.

[0125] The scheduling strategy prediction module is configured to input historical output scheduling strategies corresponding to a plurality of current time points into a preset causal neural network, so that the causal neural network predicts output scheduling strategies of the next time point corresponding to each current time point; wherein the historical output scheduling strategies include historical output scheduling values corresponding to the energy conversion system, the combined heat and power system, and the electric-to-gas system, respectively.

[0126] The multi-objective solving module is configured to solve the scheduling optimization model based on a plurality of output scheduling strategies of the next time point under the preset output constraint of the scheduling optimization model, to generate a target output scheduling strategy at the minimum operation cost.

[0127] The output control module is configured to control the output of the energy conversion system, the combined heat and power system, and the electric-to-gas system according to the target output scheduling strategy, respectively.

[0128] The generation of the causal neural network comprises:

[0129] ​The sample output scheduling strategy corresponding to each sample moment is taken as the input of the neural network, the actual output scheduling strategy of the next moment corresponding to each sample moment is taken as the output of the neural network, the neural network is iteratively trained until the neural network converges, and the trained neural network is taken as the causal neural network;

[0130] At each training time, the sample output scheduling strategy corresponding to each sample moment is input into the neural network, so that the neural network outputs the output scheduling strategy prediction result corresponding to each sample moment; the output scheduling strategy prediction result corresponding to each sample moment is compared with the actual output scheduling strategy of the next moment, and the causal coefficient in the neural network is adjusted according to the comparison result;

[0131] The causal coefficient is used to represent the causal relationship between the sample output scheduling strategy corresponding to each sample moment and the actual output scheduling strategy of the next moment.

[0132] It should be noted that the apparatus embodiments described above are merely illustrative, wherein the modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical modules, which can be located in one place or distributed on multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment. In addition, the connection relationship between the modules in the apparatus embodiment provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0133] Those skilled in the art can clearly understand that, for the convenience and brevity, the specific working process of the apparatus described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0134] On the basis of the above-mentioned various energy scheduling methods based on multi-objective model solving, the present application correspondingly provides terminal device embodiments.

[0135] An embodiment of the present application provides a terminal device, which comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements an energy scheduling method based on multi-objective model solving according to any one of the method embodiments of the present application when executing the computer program.

[0136] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server, and other computing terminal devices. The terminal device can include, but is not limited to, a processor and a memory.

[0137] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the terminal device, and connects all parts of the terminal device through various interfaces and lines.

[0138] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application required by a function, etc. The data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device or other volatile solid-state memory device.

[0139] On the basis of the above-mentioned various embodiments of the energy scheduling method based on multi-objective model solving, the application further provides a storage medium.

[0140] An embodiment of the application provides a storage medium, which comprises a stored computer program, wherein when the computer program runs, the device where the computer readable storage medium is located executes an energy scheduling method based on multi-objective model solving.

[0141] The storage medium is a computer readable storage medium, and the computer program is stored in the computer readable storage medium. When the computer program is executed by a processor, steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0142] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the scope of protection of the present application.

Claims

1. An energy scheduling method based on multi-objective model solving, characterized in that, The application relates to a scheduling optimization method and device for a combined energy conversion system, and belongs to the field of energy conversion. According to first basic data of an energy conversion system, second basic data of a combined heat and power system and third basic data of an electric-gas conversion system, a scheduling optimization model for minimizing operation cost is constructed; wherein the first basic data comprises first output data and first operation cost data of the energy conversion system; the second basic data comprises second output data and second operation cost data of the combined heat and power system; and the third basic data comprises third output data and third operation cost data of the electric-gas conversion system; A plurality of historical output scheduling strategies corresponding to current time instants are input into a preset causal neural network, so that the causal neural network predicts an output scheduling strategy of a next time instant corresponding to each current time instant; wherein the historical output scheduling strategy comprises historical output scheduling values corresponding to the energy conversion system, the combined heat and power system and the electric-gas conversion system respectively; Under preset output constraints of the scheduling optimization model, the scheduling optimization model is solved based on a plurality of output scheduling strategies of the next time instants, so as to generate a target output scheduling strategy at the time of minimizing operation cost; According to the target output scheduling strategy, the output of the energy conversion system, the combined heat and power system and the electric-gas conversion system is controlled respectively; The generation of the causal neural network comprises: During each training, the sample output scheduling strategy corresponding to each sample time instant is input into the neural network, so that the neural network outputs an output scheduling strategy prediction result corresponding to each sample time instant; the output scheduling strategy prediction result corresponding to each sample time instant is compared with the actual output scheduling strategy of the next time instant, and a causal coefficient in the neural network is adjusted according to the comparison result; The causal coefficient is used to represent the causal relationship between the sample output scheduling strategy corresponding to each sample time instant and the actual output scheduling strategy of the next time instant; The energy conversion system is a gas turbine for realizing conversion among power supply, heat supply and gas supply; the combined heat and power system is a refrigeration machine device capable of realizing integration of refrigeration, heat supply and power generation; and the electric-gas conversion system is an electric-gas conversion device for converting wind power into natural gas; The first operation cost data comprises power generation cost data of the gas turbine, carbon emission cost data of the gas turbine and energy conversion cost data of the gas turbine; the second operation cost data comprises conversion cost data of the refrigeration machine device, carbon emission cost data of the refrigeration machine device and energy conversion cost data of the refrigeration machine device; and the third operation cost data comprises conversion cost data of the electric-gas conversion device into natural gas, carbon emission cost data of the electric-gas conversion device and energy conversion cost data of the electric-gas conversion device; The scheduling optimization model for minimizing operation cost is constructed according to the following formula: The preset output constraints of the scheduling optimization model comprise: ； ； ； ； ； ； ； ； ； ； ； ； wherein, a dispatch optimization model for representing minimization of total operating cost of the gas turbine, the chiller plant, and the electric-to-gas plant, a total start-up cost of the gas turbine, the chiller plant, and the electric-to-gas plant, a total carbon emission cost of the gas turbine, the chiller plant, and the electric-to-gas plant, a total energy conversion cost of the gas turbine, the chiller plant, and the electric-to-gas plant; a generation cost of the gas turbine; a conversion cost of the chiller plant; a generation cost of the chiller plant; a conversion cost of the electric-to-gas plant to natural gas; a carbon emission cost of the gas turbine; a carbon emission cost of the chiller plant; a carbon emission cost of the electric-to-gas plant; an energy conversion cost of the gas turbine operation; an energy conversion cost of the combined heat and power system; an energy conversion cost of the electric-to-gas plant; , b and c are different efficiency coefficients of micro gas turbine; P mt is the output power of gas turbine; is the input thermal power of refrigeration equipment; is the refrigeration coefficient of refrigeration equipment; D ac is the conversion coefficient of refrigeration by electric energy of refrigeration equipment; P ac is the operation energy consumption of refrigeration equipment; is the conversion efficiency of electric-gas equipment; is the power of electric-gas technology; α, β, γ are respectively quadratic factor, linear factor and constant factor of carbon emission of gas turbine; r ac is the carbon emission factor of refrigeration equipment; is the energy conversion factor of refrigeration equipment; is the energy conversion factor of electric-gas equipment.

2. The energy dispatching method based on multi-objective model solving of claim 1, wherein, ​ ； ； ； wherein is a preset upper limit value for the output power of the gas turbine, is a preset upper limit value for the input thermal power of the chiller plant, is a preset upper limit value for the power of the power-to-gas plant.

3. The energy dispatching method based on multi-objective model solving of claim 2, wherein, The neural network comprises a causal selection network and a causal prediction network. The inputting of the sample output scheduling strategy corresponding to each sample moment into the neural network comprises: The inputting of the sample output scheduling strategy corresponding to each sample moment into the neural network comprises: The causal prediction network predicts the output scheduling strategy prediction result corresponding to each sample moment according to the strategy feature.

4. The energy dispatching method based on multi-objective model solving of claim 3, wherein, The comparing of the output scheduling strategy prediction result corresponding to each sample moment with the actual output scheduling strategy of the next moment and the adjustment of the causal coefficient in the neural network according to the comparison result comprise: The comparing of the output scheduling strategy prediction result corresponding to each sample moment with the actual output scheduling strategy of the next moment and the adjustment of the causal coefficient in the neural network according to the comparison result comprise: ； wherein, representing a loss value, is the output scheduling strategy prediction result corresponding to the sample moment; is the actual output scheduling strategy at the next moment; m is the data size of the small batch sampling, is the sample index of the small batch sampling; The calculation of the gradient value of the output scheduling strategy prediction result corresponding to each sample moment according to the updated loss function comprises: The adjustment of the causal coefficient in the neural network according to the gradient value corresponding to each output scheduling strategy prediction result comprises:

5. The energy dispatching method based on multi-objective model solving of claim 4, wherein, The calculation of the gradient value of the output scheduling strategy prediction result corresponding to each sample moment according to the updated loss function comprises: The calculation of the loss value of each output scheduling strategy prediction result according to the loss function; The multiplication of the loss value of each output scheduling strategy prediction result with the activation function of the causal selection network and the causal prediction network respectively to obtain the gradient value corresponding to each output scheduling strategy prediction result in the causal selection network and the causal prediction network respectively.

6. The energy dispatching method based on multi-objective model solving of claim 5, wherein, The adjustment of the causal coefficient in the neural network according to the gradient value corresponding to each output scheduling strategy prediction result comprises: The adjustment of the causal coefficient in the neural network according to the gradient value corresponding to each output scheduling strategy prediction result comprises:

7. An energy scheduling device based on multi-objective model solving, characterized in that, The model construction module is configured to construct a scheduling optimization model for minimizing operation cost according to first basic data of an energy conversion system, second basic data of a combined cooling heating and power system, and third basic data of an electric-to-gas system; the first basic data comprises first output data and first operation cost data of the energy conversion system; the second basic data comprises second output data and second operation cost data of the combined cooling heating and power system; and the third basic data comprises third output data and third operation cost data of the electric-to-gas system. The model construction module is configured to construct a scheduling optimization model for minimizing operation cost according to first basic data of an energy conversion system, second basic data of a combined cooling heating and power system, and third basic data of an electric-to-gas system; the first basic data comprises first output data and first operation cost data of the energy conversion system; the second basic data comprises second output data and second operation cost data of the combined cooling heating and power system; and the third basic data comprises third output data and third operation cost data of the electric-to-gas system. ​ The scheduling strategy prediction module is configured to input historical output scheduling strategies corresponding to a plurality of current time points into a preset causal neural network, so that the causal neural network predicts an output scheduling strategy of a next time point corresponding to each current time point; wherein the historical output scheduling strategies include historical output scheduling values corresponding to the energy conversion system, the combined cooling heating and power system, and the electric-to-gas system respectively; The multi-objective solving module is configured to solve the scheduling optimization model based on the output scheduling strategies of the plurality of next time points under preset output constraints of the scheduling optimization model, to generate a target output scheduling strategy with the minimum operation cost; The output control module is configured to control the output of the energy conversion system, the combined cooling heating and power system, and the electric-to-gas system respectively according to the target output scheduling strategy. The generation of the causal neural network includes: iteratively training the neural network with each sample output scheduling strategy corresponding to a sample time point as an input of the neural network and an actual output scheduling strategy of a next time point corresponding to each sample time point as an output of the neural network, until the neural network converges, and taking the trained neural network as the causal neural network; In each training, each sample output scheduling strategy corresponding to a sample time point is input into the neural network, so that the neural network outputs an output scheduling strategy prediction result corresponding to each sample time point; the output scheduling strategy prediction result corresponding to each sample time point is compared with an actual output scheduling strategy of a next time point corresponding thereto, and a causal coefficient in the neural network is adjusted according to a comparison result; The causal coefficient is used to represent a causal relationship between the sample output scheduling strategy corresponding to the sample time point and the actual output scheduling strategy of the next time point corresponding thereto. The energy conversion system is a gas turbine for realizing conversion between power supply, heat supply, and gas supply; the combined cooling heating and power system is a refrigeration machine device capable of realizing integration of refrigeration, heat supply, and power generation; and the electric-to-gas system is an electric-to-gas device for converting wind power into natural gas. The first operation cost data includes power generation cost data of the gas turbine, carbon emission cost data of the gas turbine, and energy conversion cost data of the gas turbine; the second operation cost data includes conversion cost data of the refrigeration machine device, carbon emission cost data of the refrigeration machine device, and energy conversion cost data of the refrigeration machine device; and the third operation cost data includes cost data of the electric-to-gas device for converting into natural gas, carbon emission cost data of the electric-to-gas device, and energy conversion cost data of the electric-to-gas device. The scheduling optimization model for minimizing the operation cost is constructed according to the following formula: ； ； ； ； ； ； ； ； ； ； ； ； wherein, for representing the total operating cost of the gas turbine, chiller plant and electric-to-gas plant, for representing the total start-up cost of the gas turbine, chiller plant and electric-to-gas plant, for representing the total carbon emission cost of the gas turbine, chiller plant and electric-to-gas plant, for representing the total energy conversion cost of the gas turbine, chiller plant and electric-to-gas plant; for the generation cost of the gas turbine; for the conversion cost of the chiller plant; for the generation cost of the chiller plant; for the conversion cost of the electric-to-gas plant to natural gas; for the carbon emission cost of the gas turbine; for the carbon emission cost of the chiller plant; for the carbon emission cost of the electric-to-gas plant; for the energy conversion cost of the gas turbine operation; for the energy conversion cost of the combined heat and power system; for the energy conversion cost of the electric-to-gas plant; , b and c are different efficiency coefficients of micro gas turbines; P mt is the output power of the gas turbine; is the input thermal power of the refrigeration machine device; is the refrigeration coefficient of the refrigeration machine device; D ac is the conversion coefficient of the refrigeration machine through electric energy; P ac is the operating energy consumption of the refrigeration machine device; is the conversion efficiency of the electric-to-gas device; is the power of the electric-to-gas technology; α, β, γ are the quadratic factor, linear factor and constant factor of carbon emissions of the gas turbine, respectively; r ac is the carbon emission factor of the refrigeration machine device; is the energy conversion factor of the refrigeration machine device; is the energy conversion factor of the electric-to-gas device.

8. A terminal device, comprising: The processor executes the computer program to implement the energy scheduling method based on the multi-objective model solving according to any one of claims 1 to 6. The processor executes the computer program to implement the energy scheduling method based on the multi-objective model solving according to any one of claims 1 to 6.

9. A storage medium, characterized by The storage medium comprises a stored computer program, wherein the computer program controls a device where the storage medium is located to perform the energy scheduling method based on multi-target model solving according to any one of claims 1 to 6 when the computer program is running.

Images