Method and system for determining capacity configuration in integrated energy system and electronic equipment

Abstract

The invention provides a method and system for determining capacity configuration in an integrated energy system and electronic equipment, and the method comprises the steps: S1, constructing an alternative equipment capacity population comprising different first capacity combinations through employing a genetic algorithm based on equipment correlation parameters; s2, inputting a to-be-predicted time period and the alternative equipment capacity population into the energy storage start-stop state prediction model, and predicting the corresponding actual energy storage start-stop operation state of each energy equipment under each target time sequence boundary in different first capacity combinations; s3, calculating the fitness of the first capacity combination by adopting a fitness function of a genetic algorithm; s4, performing iterative updating on the alternative equipment capacity population according to the fitness, and returning to execute the step S2 until an iteration stop condition is reached; and S5, obtaining the target capacity combination with the minimum fitness, and obtaining the comprehensive energy system. According to the method, MILP solution is converted into LP through integer variable state prediction, the online calculation complexity is remarkably reduced, and the solution efficiency is improved.

Description

Technical Field

[0001] This disclosure relates to the field of integrated energy system planning technology, and in particular to a method, system, and electronic equipment for determining capacity configuration in an integrated energy system. Background Technology

[0002] Integrated Energy Systems (IES) coordinate multiple energy sources such as electricity, heat, cooling, and natural gas to achieve energy complementarity, improve energy efficiency, and reduce system operating costs. They are a crucial supporting technology for promoting energy transition and achieving carbon neutrality. In recent years, with the large-scale integration of fluctuating renewable energy sources such as wind and solar power, the complexity of integrated energy system planning and operation optimization has increased significantly.

[0003] Currently, mixed-integer linear programming (MILP) is widely used for modeling and solving integrated energy system optimization problems. The advantage of MILP lies in its ability to accurately describe the start-up and shutdown, operational constraints of various energy devices, and the coupling relationships between multiple energy flows. However, because MILP simultaneously involves continuous and integer decision variables, it is typically an NP-hard (nondeterministic polynomial) problem. In practical applications, the computational complexity increases exponentially with the scale of the system, resulting in low solution efficiency and long processing times, making it difficult to meet the stringent requirements for computational speed and real-time performance in planning, design, and real-time operation control.

[0004] To improve the efficiency of solving MILP problems, current research has attempted to reduce computational complexity through mathematical decomposition, heuristic algorithms, and approximation algorithms. However, while traditional mathematical decomposition methods (such as Benders decomposition, an algorithm for solving mixed integer programming problems) can reduce complexity, they require reconstructing the model structure, which is difficult to implement in practical engineering. Furthermore, they are prone to information loss leading to a decline in solution quality, making it difficult to balance efficiency and accuracy. Traditional heuristic algorithms (such as genetic algorithms and particle swarm optimization) and approximation strategies (such as continuous relaxation) can improve the solution speed, but they rely on empirical parameter tuning and lack in-depth modeling of the system's physical characteristics, resulting in poor stability, insufficient generalization ability, and limited application scope. Approximation methods may produce suboptimal solutions, affecting the overall economy and security of integrated energy systems.

[0005] Existing MILP solution methods suffer from problems such as the trade-off between computational efficiency and accuracy, insufficient algorithm generalization ability, and difficulty in implementation in practical engineering, which seriously hinder the development of large-scale integrated energy system optimization planning and real-time scheduling technology. Summary of the Invention

[0006] The technical problem to be solved by this disclosure is to overcome the shortcomings of low efficiency and difficulty in balancing accuracy and efficiency in solving mixed-integer linear programming in existing integrated energy systems, and to provide a method, system and electronic equipment for determining capacity configuration in integrated energy systems.

[0007] This disclosure solves the above-mentioned technical problems through the following technical solution:

[0008] According to a first aspect of this disclosure, a method for determining capacity configuration in an integrated energy system is provided, the method comprising:

[0009] S1. Based on the equipment association parameters of multiple energy devices in the system to be constructed, a genetic algorithm is used to construct a population of candidate equipment capacities including different combinations of first capacities.

[0010] S2. Input the time period to be predicted and the capacity population of the candidate equipment into the energy storage start-up and shutdown state prediction model obtained by pre-offline training, and predict the actual energy storage start-up and shutdown operation state of each energy device under each target time boundary in different first capacity combinations.

[0011] Wherein, the target time series boundary is obtained based on the time period to be predicted, and the actual energy storage start-stop operation status is a characterization parameter for whether the energy equipment is in operation.

[0012] S3. Based on the actual energy storage start-up and shutdown operation states corresponding to different first capacity combinations, the fitness of different first capacity combinations is calculated using the fitness function of the genetic algorithm.

[0013] S4. Iteratively update the candidate equipment capacity population according to the fitness, and return to step S2 until the genetic algorithm meets the convergence condition or the fitness reaches the iteration stop condition.

[0014] S5. Obtain the first capacity combination with the smallest fitness as the target capacity combination of the system to be constructed, and obtain the integrated energy system.

[0015] Optionally, the training process of the energy storage start-stop state prediction model includes:

[0016] S01. Obtain historical operation data, and perform typical day clustering processing on the historical operation data to construct a training sample set;

[0017] Among them, different typical days correspond to different sample time series boundaries;

[0018] S02. Based on the pre-configured multiple sample capacity combinations and the training sample set, an optimization solver is used to solve the sample energy storage start-up and shutdown states of the energy device in each sample capacity combination under different sample time-series boundaries.

[0019] S03. Based on the training sample set, the sample capacity combination, and the corresponding sample energy storage start-stop state, perform model training on the preset model to obtain the energy storage start-stop state prediction model.

[0020] Optionally, step S02 includes:

[0021] S021. Normalize the configuration capacity of each energy device in each sample capacity combination to obtain the sample normalized capacity value of each energy device;

[0022] S022. Discretize the sample per-unit capacity value based on a preset step size to obtain the sample discrete capacity value.

[0023] S023. Using the discrete capacity value of each energy device in the sample capacity combination, the training sample set, and the candidate energy storage start-stop state as input to the optimization solver, solve for the sample energy storage start-stop state of each energy device in the sample capacity combination under different sample time-series boundaries.

[0024] In this context, the energy storage start / stop status is represented by 0 to indicate that the energy equipment is not running, and by 1 to indicate that the energy equipment is running.

[0025] Optionally, step S023 can be implemented using the following formula:

[0026]

[0027] in, This indicates the start / stop status of the sample energy storage corresponding to each of the aforementioned energy devices. In the function The set of values ​​for the independent variable u when the minimum value is obtained, where C represents the sample size combination, u represents the start-stop state of the alternative energy storage, A and B represent weight coefficients, and d is the vector coefficient.

[0028] Optionally, step S03 includes:

[0029] The discrete capacity value of each energy device in the training sample set and the sample capacity combination is used as the input of the preset model, and the corresponding sample energy storage start-up and shutdown state is used as the output of the preset model. The preset model is trained to obtain the energy storage start-up and shutdown state prediction model.

[0030] Optionally, the preset model includes the XGBoost (an optimized distributed gradient boosting library) model.

[0031] Optionally, step S2 includes:

[0032] S21. Normalize the configuration capacity of each energy device in the first capacity combination to obtain a first normalized capacity value for each energy device;

[0033] S22. Discretize the first per-unit capacity value based on a preset step size to obtain the first discrete capacity value of the first capacity combination.

[0034] S23. Input the first discrete capacity value of the time period to be predicted and the first capacity combination into the energy storage start-up and shutdown state prediction model to predict the actual energy storage start-up and shutdown operation state corresponding to different first capacity combinations.

[0035] Optionally, step S3 includes:

[0036] S31. Based on the actual start-up and shutdown operation status of each first capacity combination, calculate the operating cost of each first capacity combination in the predicted time period in parallel.

[0037] S32. Obtain the investment cost for each of the first capacity combinations;

[0038] S33. Using the fitness function of the genetic algorithm, the fitness of each first capacity combination is calculated based on the investment cost and the operating cost.

[0039] Optionally, step S4 includes:

[0040] S41. Arrange the different combinations of the first capacity in ascending order of fitness;

[0041] S42. A preset number of the first capacity combinations are used as parent individuals. Selection, crossover, and mutation operations are performed on the parent individuals in sequence to generate a new generation of candidate device capacity population composed of new different first capacity combinations.

[0042] According to a second aspect of this disclosure, a system for determining capacity configuration in an integrated energy system is provided, the system comprising a construction module, a prediction module, a calculation module, an iteration module, and a determination module;

[0043] The construction module is used to construct a population of candidate device capacities, including different combinations of first capacities, based on the device association parameters of multiple energy devices in the system to be constructed, using a genetic algorithm.

[0044] The prediction module is used to input the time period to be predicted and the capacity population of the candidate equipment into the energy storage start-up and shutdown state prediction model obtained by pre-offline training, and to predict the actual energy storage start-up and shutdown operation state of each energy device under each target time boundary in different first capacity combinations.

[0045] Wherein, the target time series boundary is obtained based on the time period to be predicted, and the actual energy storage start-stop operation status is a characterization parameter for whether the energy equipment is in operation.

[0046] The calculation module is used to calculate the fitness of different first capacity combinations based on the actual energy storage start-up and shutdown operation states corresponding to different first capacity combinations, using the fitness function of the genetic algorithm.

[0047] The iteration module is used to iteratively update the candidate equipment capacity population according to the fitness, and return to call the prediction module until the genetic algorithm meets the convergence condition or the fitness reaches the iteration stopping condition.

[0048] The determining module is used to obtain the first capacity combination with the smallest fitness as the target capacity combination of the system to be built, thereby obtaining the integrated energy system.

[0049] According to a third aspect of this disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and for running on the processor, wherein the processor executes the computer program to implement the determination method described in the first aspect of this disclosure.

[0050] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the determination method described in the first aspect of this disclosure.

[0051] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the determination method described in the first aspect of this disclosure.

[0052] Based on common knowledge in the field, the above optional conditions can be combined arbitrarily to obtain the optional examples of this disclosure.

[0053] The positive improvements of this disclosure are as follows: In the offline stage, a perfect policy set is used to train the energy storage start-up and shutdown state prediction model. In the online stage, the MILP solution is transformed into LP (linear programming) or low-dimensional constraint optimization through integer variable state prediction. This maintains the approximate optimality of the policy, significantly reduces the online computational complexity, and greatly improves the solution efficiency. At the same time, the input of the energy storage start-up and shutdown state prediction model includes the equipment association parameters of energy equipment such as wind power, photovoltaic, load, and electricity price at future N time points, as well as the per-unit capacity vector of energy equipment. After training in multiple scenarios and configuration combinations, it has good generalization ability across seasons and energy equipment combinations, and is suitable for different operating conditions. Attached Figure Description

[0054] Figure 1 This is a first flowchart of the method for determining capacity configuration in an integrated energy system according to Embodiment 1 of this disclosure;

[0055] Figure 2 This is a topology diagram of the integrated energy system according to Embodiment 1 of this disclosure;

[0056] Figure 3 This is a second flowchart of the method for determining capacity configuration in an integrated energy system according to Embodiment 1 of this disclosure;

[0057] Figure 4 This is a third flowchart of the method for determining capacity configuration in an integrated energy system according to Embodiment 1 of this disclosure;

[0058] Figure 5 This is the fourth flowchart of the method for determining capacity configuration in the integrated energy system of Embodiment 1 of this disclosure;

[0059] Figure 6 This is the fifth flowchart of the method for determining capacity configuration in the integrated energy system of Embodiment 1 of this disclosure;

[0060] Figure 7 This is a schematic diagram of the system module for determining capacity configuration in the integrated energy system of Embodiment 2 of this disclosure;

[0061] Figure 8 This is a schematic diagram of the structure of an electronic device according to Embodiment 3 of this disclosure. Detailed Implementation

[0062] The present disclosure is further illustrated below by way of embodiments, but the present disclosure is not limited to the scope of the embodiments described herein.

[0063] The prefixes such as "first" and "second" used in this disclosure are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this disclosure does not constitute a limitation on the described objects. The description of the described objects is given in the context of the embodiments, and the use of such prefixes should not constitute unnecessary restrictions. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0064] Example 1

[0065] In one specific embodiment, a method for determining the capacity configuration in an integrated energy system is provided, such as... Figure 1 As shown, the determination method includes:

[0066] S1. Based on the equipment association parameters of multiple energy devices in the system to be constructed, a genetic algorithm is used to construct a population of candidate equipment capacities including different combinations of first capacities.

[0067] S2. Input the time period to be predicted and the capacity population of the alternative equipment into the energy storage start-up and shutdown state prediction model obtained by pre-offline training, and predict the actual energy storage start-up and shutdown operation state of each energy device under each target time boundary in different first capacity combinations.

[0068] Among them, the target time series boundary is obtained based on the period to be predicted, and the actual energy storage start-up and shutdown operation status is a characterization parameter for whether the energy equipment is in operation.

[0069] S3. Based on the actual start-up and shutdown operation states of different first capacity combinations, the fitness of different first capacity combinations is calculated using the fitness function of a genetic algorithm.

[0070] S4. Iteratively update the candidate equipment capacity population based on fitness, and return to step S2 until the genetic algorithm meets the convergence condition or the fitness reaches the iteration stopping condition.

[0071] S5. Obtain the first capacity combination with the minimum fitness as the target capacity combination of the system to be built, and obtain the integrated energy system.

[0072] Specifically, such as Figure 2 As shown, an integrated energy system includes multiple energy sources such as photovoltaics, wind turbines, batteries, and the power grid, all working together to provide electricity load to users. Therefore, it is necessary to determine the capacity configuration of each energy device in the integrated energy system to achieve a supply-demand balance with the electricity load. Determining the capacity configuration of the integrated energy system can be transformed into a two-layer solution process: upper-level capacity optimization and lower-level operation optimization.

[0073] When determining the capacity configuration of the integrated energy system, the equipment association parameters of each energy device in the system to be constructed are first obtained online through step S1. These parameters include, for example, the maximum wind power installed capacity, the maximum photovoltaic installed capacity, the battery charging and discharging power, and the power purchase and sale capacity of the power grid. These parameters serve as boundary conditions for the capacity configuration of the integrated energy system. Then, a genetic algorithm is used to treat the configuration capacity corresponding to each energy device as a chromosome. Within the range of the equipment association parameters, several different first capacity combinations composed of different configuration capacities of each energy device are constructed as a candidate equipment capacity population.

[0074] After obtaining the candidate equipment capacity population, step S2 performs lower-level operation optimization. The time period to be predicted and the candidate equipment capacity population are input into the pre-trained offline energy storage start-up and shutdown state prediction model. The energy storage start-up and shutdown state prediction model is used to predict the actual energy storage start-up and shutdown operation state of each energy device under the configured capacity of each first capacity combination within the target time boundary corresponding to the time period to be predicted. Different target time boundaries correspond to different typical daily operation scenarios, such as the power generation period of photovoltaic energy equipment in different seasons of the next year, and the power purchased / sold from the grid in different seasons.

[0075] The energy storage start-up and shutdown state prediction model can be obtained in advance through offline training, and the actual start-up and shutdown operation states of each energy device under different target time-series boundaries can be output using [0,1] variables. For example, when the output is 0 under the first time-series boundary, it indicates that the energy device is in a stopped operation state under the first time-series boundary; when the output is 1 under the second time-series boundary, it indicates that the energy device is in a started operation state under the first time-series boundary. In this way, the continuous variables of energy storage start-up and shutdown of the energy device can be converted into an integer state sequence, so that the MILP solution is transformed into LP or low-dimensional constraint optimization.

[0076] Step S3: Based on the prediction results of the actual energy storage start-up and shutdown operation status for each first capacity combination using the energy storage start-up and shutdown state prediction model, the fitness of each first capacity combination is calculated using the fitness function of the genetic algorithm, so as to achieve the evaluation of each first capacity combination.

[0077] Then, in step S4, parent individuals are selected based on the fitness of each first capacity combination to generate a new generation of candidate equipment capacity population. Then, step S2 is returned to predict the actual energy storage start-up and shutdown operation status of each new first capacity combination in the new generation of candidate equipment capacity population until the genetic algorithm meets the convergence condition or the fitness reaches the iteration stopping condition.

[0078] By continuously generating new candidate equipment capacity populations as the boundary of the lower-level operation optimization, the optimization results of the upper-level capacity optimization are passed to the lower-level operation optimization for further refinement of the capacity boundary conditions. If the lower-level operation optimization finds that the current candidate equipment capacity population performs poorly in terms of operating costs, it feeds back to the upper-level capacity optimization by adjusting equipment configurations and parameters for further optimization. Through iterative optimization, the upper-level capacity optimization and the lower-level operation optimization alternate operation, eventually converging to the global optimal solution.

[0079] Step S5 selects the first capacity combination with the lowest fitness from the latest generation of candidate equipment capacity population as the target capacity combination for deployment as the configuration capacity of each energy device in the integrated energy system, which satisfies the supply and demand balance of electricity load and ensures the optimal capacity configuration of each energy device in the integrated energy system.

[0080] In this specific embodiment, a perfect policy set is used to train the energy storage start-up and shutdown state prediction model in the offline stage. In the online stage, the MILP solution is transformed into LP or low-dimensional constraint optimization through integer variable state prediction. This maintains the approximate optimality of the policy, significantly reduces the online computational complexity, and greatly improves the solution efficiency.

[0081] In one specific implementation, such as Figure 3 As shown, the training process of the energy storage start-stop state prediction model includes:

[0082] S01. Obtain historical operation data and perform typical daily clustering processing on the historical operation data to construct a training sample set;

[0083] Among them, different typical days correspond to different sample time series boundaries;

[0084] S02. Based on the pre-configured multiple sample capacity combinations and training sample sets, use an optimization solver to solve the sample energy storage start-up and shutdown states of the energy devices in each sample capacity combination under different sample time-series boundaries.

[0085] S03. Based on the training sample set, sample capacity combination and corresponding sample energy storage start-stop state, train the preset model to obtain the energy storage start-stop state prediction model.

[0086] Specifically, it can obtain daily historical operation data within a preset time period (e.g., the past year), and perform typical day clustering on daily historical operation data across seasons and months to obtain K typical days. K is a positive integer. By merging weather types with similar operational conditions, the amount of data processing is effectively reduced, and data processing efficiency is improved.

[0087] Each typical day is used as a sample time series boundary, and historical operational data under the same sample time series boundary are fused to obtain operational data for typical days corresponding to different sample time series boundaries. The training sample set constitutes This ensures coverage of various operating conditions across different scenarios.

[0088]

[0089] Construct a set of energy equipment parameters As external boundary conditions, multiple different combinations of sample sizes are constructed based on the set of external boundary conditions. For example, within the range of external boundary conditions, multiple different combinations of sample sizes are obtained through exhaustive search. ,in, This represents the configuration capacity of the nth type of energy equipment, where n is a positive integer.

[0090] For different combinations of sample sizes Typical daily operational data corresponding to different sample time series boundaries The optimal operating strategy for each sample size combination C under different sample time boundaries is solved using a mathematical optimization solver. :

[0091]

[0092] in, This indicates the start / stop status of the sample energy storage for each energy device. In the function The set of values ​​for the independent variable u when the minimum value is obtained, where C represents the sample size combination, u represents the start-stop state of the alternative energy storage, A and B represent weight coefficients, and d is the vector coefficient.

[0093] This leads to a perfect strategy set. The perfect strategy set Operational strategies for different sample size combinations under different sample time boundaries This is converted into the sample energy storage start-up and shutdown states of different energy devices under different sample time boundaries. t represents the predicted time.

[0094] Combine training sample set and sample size Sample size combination as input to the preset model Sample energy storage start-up and shutdown status of various energy devices under different sample time boundaries The energy storage start-up and shutdown state prediction model can be obtained by training the preset model as its output. The preset model can be an XGBoost classification model.

[0095] In one specific embodiment, step S02 includes:

[0096] S021. Normalize the configuration capacity of each energy device in each sample capacity combination to obtain the sample normalized capacity value of each energy device.

[0097] S022. Discretize the sample per-unit capacity value based on the preset step size to obtain the sample discrete capacity value.

[0098] S023. Take the discrete capacity value of each energy device in the sample capacity combination, the training sample set and the candidate energy storage start-up and shutdown state as the input of the optimization solver, and solve the sample energy storage start-up and shutdown state of each energy device in the sample capacity combination under different sample time-series boundaries.

[0099] In this context, the energy storage start-stop status is represented by 0 to indicate that the energy equipment is not running, and 1 to indicate that the energy equipment is running.

[0100] Specifically, to ensure the multi-scenario adaptability of the energy storage start-up and shutdown state prediction model, we can first analyze the configuration capacity corresponding to different energy devices in each sample capacity combination C. Based on system maximum load Per-unit scaling yields the sample per-unit capacity value for each energy device. :

[0101]

[0102] with preset step size Per-unit capacity of samples Discretization is performed to obtain a sample capacity combination consisting of the discrete sample capacity values ​​corresponding to the energy devices. For example, with Discretization is performed to obtain .

[0103] To avoid dimensional expansion caused by the Cartesian product, a constraint can be imposed using the sum of capacities:

[0104]

[0105] Specifically, the per-unit capacity of samples can be ensured through a combination of logical mutual exclusion and collaborative rule pruning. The effectiveness.

[0106] Next, combine the sample sizes. The discrete capacity value of each energy device, the training sample set, and the start-up and shutdown states of the candidate energy storage are used as inputs to the optimization solver. The mathematical optimization solver is then used to solve for each sample capacity combination. Optimal running strategy under different sample time boundaries Each sample size combination The energy storage start-up and shutdown states of the energy devices in the sample under different sample time boundaries .

[0107] In one specific implementation, step S03 includes: using the discrete capacity value of each energy device in the training sample set and sample capacity combination as the input of the preset model, and the corresponding sample energy storage start-up and shutdown state as the output of the preset model, training the preset model to obtain the energy storage start-up and shutdown state prediction model.

[0108] Specifically, after obtaining the combination of discretized sample sizes And the corresponding sample energy storage start / stop status. Then, combine the sample sizes. Integrate with temporal boundary conditions into a global boundary Input a preset model, where This represents the solution boundary time series information from time t to time t+N, and the sample energy storage start-up and shutdown states of different energy devices under different sample time series boundaries. As the output of the preset model, the preset model is trained to obtain the energy storage start-up and shutdown state prediction model. , This indicates the start-up and shutdown operation status of energy equipment predicted by the energy storage start-up and shutdown status prediction model. The parameter is The function.

[0109] In one specific embodiment, step S2 includes:

[0110] S21. Normalize the configured capacity of each energy device in the first capacity combination to obtain the first normalized capacity value of each energy device;

[0111] S22. Discretize the first per-unit capacity value based on a preset step size to obtain the first discrete capacity value of the first capacity combination.

[0112] S23. Input the first discrete capacity value of the time period to be predicted and the first capacity combination into the energy storage start-up and shutdown state prediction model to predict the actual energy storage start-up and shutdown operation state corresponding to different first capacity combinations.

[0113] Specifically, when using the energy storage start-up and shutdown state prediction model to predict the actual start-up and shutdown operation status of different energy devices in the first capacity combination under different target time boundaries, it is also necessary to standardize and discretize the configuration capacity corresponding to different energy devices in the first capacity combination to obtain the first discrete capacity value of the first capacity combination. Then, the first discrete capacity value of the first capacity combination and the time period to be predicted are input into the energy storage start-up and shutdown state prediction model for prediction.

[0114] This method employs clustering with historical operational data to determine different typical days, dividing the prediction period into several target time-series boundaries. Once the target time-series boundaries are determined, their corresponding typical daily operational data are also determined. Then, based on the first discrete capacity value of the first capacity combination and the typical daily operational data corresponding to the target time-series boundaries, the actual energy storage start-up and shutdown operation status of each energy device under different first capacity combinations can be predicted. The specific standardization and discretization processes can be found in the aforementioned detailed implementation method and will not be repeated here.

[0115] In one specific implementation, such as Figure 4 As shown, step S3 includes:

[0116] S31. Based on the actual start-up and shutdown operation status of each first capacity combination, calculate the operating cost of each first capacity combination in parallel during the period to be predicted.

[0117] S32. Obtain the investment cost for each first capacity combination;

[0118] S33. Using the fitness function of the genetic algorithm, the fitness of each first capacity combination is calculated based on the investment cost and operating cost.

[0119] Specifically, a fitness function for a genetic algorithm can be constructed based on the operating and investment costs of each first capacity combination. A combinatorial computational method based on population optimization strategies can be used to evaluate the long-term operating costs of each first capacity combination, including fuel, depreciation costs of cost equipment, maintenance costs, and energy trading costs (e.g., electricity market transaction fees).

[0120] For each first capacity combination C, the hourly boundary conditions at 8760 hours... Below, a pre-trained energy storage start-up and shutdown state prediction model is used to quickly predict the actual start-up and shutdown operation states of different energy devices under different target time-series boundaries, resulting in an integer state sequence. The operating cost of the first capacity combination C during the forecast period is calculated. Taking a forecast period of 8760 hours as an example, the formula for calculating the operating cost of the first capacity combination C during the forecast period is as follows:

[0121]

[0122] in, This includes the costs of purchasing, selling, and maintaining electricity at time t.

[0123] The calculated operating cost of the first capacity combination C is fed back to the upper-level capacity optimization. The upper-level capacity optimization combines the investment cost and operating cost of energy equipment and uses the fitness function of a genetic algorithm to calculate the fitness of each first capacity combination. Since individuals (first capacity combinations) in the same population are independent of each other, the fitness of each first capacity combination can be calculated in parallel to accelerate the evaluation speed of the first capacity combinations.

[0124] Let the candidate capacity population configuration for the kth generation be as follows: Its fitness function is:

[0125]

[0126] Where: I represents the investment cost, This represents the estimated operating cost corresponding to the first capacity combination.

[0127] In one specific embodiment, step S4 includes:

[0128] S41. Arrange the different first capacity combinations in ascending order of fitness.

[0129] S42. A first capacity combination with a preset number of individuals is used as the parent individuals. Selection, crossover and mutation operations are performed on the parent individuals in sequence to generate a new generation of candidate device capacity population composed of new different first capacity combinations.

[0130] Specifically, after calculating the fitness of each first capacity combination, the first capacity combinations are arranged in ascending order of fitness. Several first capacity combinations with lower fitness are selected as parent individuals. For example, using a multi-objective evolutionary algorithm such as NSGA-II (multi-objective genetic algorithm), selection, crossover, and mutation operations are sequentially performed on the parent individuals to generate a new generation of candidate device capacity population composed of different first capacities. This new generation of candidate device capacity population is then passed to the lower-level operation optimization for operation cost calculation. This process is iterated until the fitness function J no longer decreases significantly, or the number of iterations reaches the preset number of iterations. When the iteration stops, the first capacity combination with the lowest fitness is output as the target capacity combination configuration scheme. Deployment is carried out.

[0131] In a specific example, such as Figure 5As shown, in the offline phase, historical operating data such as wind power, photovoltaics, load, and electricity price within a preset time period are clustered by typical days of each month to obtain sample time series boundaries and typical day operating data corresponding to different typical days; the configuration capacity of candidate energy equipment is normalized according to the maximum load to obtain the configuration capacity corresponding to different energy equipment in each sample capacity combination; then, based on the typical day operating data corresponding to different sample time series boundaries and the configuration capacity corresponding to different energy equipment in each sample capacity combination, a full combination boundary condition including different equipment capacity ratios and different wind and solar load time series boundaries is generated.

[0132] Next, an optimization solver is used to solve the sample energy storage start-up and shutdown state for each sample capacity combination under the full combination boundary conditions. Based on the sample capacity combination, typical daily operation data and the corresponding sample energy storage start-up and shutdown state, XGBoost is trained. The operation strategy is solidified through machine learning or deep learning methods to obtain the energy storage start-up and shutdown state prediction model.

[0133] When determining the capacity configuration of an online integrated energy system, the associated parameters of each energy device in the current integrated energy system are obtained, such as boundary information of wind power, photovoltaic, load, and electricity price. A genetic algorithm is used to generate a population of candidate equipment capacity including multiple different first capacity combinations based on the boundary information. The time period to be predicted and the first capacity combination are input into an offline-trained energy storage start-up and shutdown state prediction model to predict the actual energy storage start-up and shutdown operation state. The MILP solution is transformed into LP or low-dimensional constraint optimization through integer variable state prediction, which accelerates the calculation of the operating cost of each first capacity combination. Then, the investment cost is added to determine the fitness. A new generation of candidate equipment capacity population is generated iteratively until the fitness no longer decreases significantly or the number of iterations reaches the preset number of iterations, and the target capacity combination with the minimum fitness is output.

[0134] Among them, such as Figure 6 As shown, the genetic algorithm first randomly generates a population of candidate device capacities including multiple first capacity combinations. It then uses parallel computing to calculate the operating cost of each first capacity combination. By accumulating the costs, it performs fitness assessment. If the fitness decreases, it generates a new generation of candidate device capacity populations through selection, crossover, and mutation. It then uses parallel computing again to calculate the operating cost of each first capacity combination in the new generation of candidate device capacity populations. This process is iterated until the fitness no longer decreases significantly or the number of iterations reaches the preset number of iterations, at which point the target capacity combination is output.

[0135] In a specific example, battery-related variables include: charging power. (KW (kilowatts)), discharge power (KW), Charging Status Discharge state State of charge (%); Wind power-related variables include: wind power output (KW), wind power installed capacity (KW); photovoltaic-related variables include: photovoltaic output (KW), photovoltaic installed capacity (KW); Grid-related variables include: purchased power. (KW), Power Sales (KW).

[0136] The constraints are as follows:

[0137] ;

[0138] ;

[0139] ;

[0140] ;

[0141] ;

[0142] The objective function is:

[0143] Using 8760 hours of annual time-series data on load, electricity price, wind speed, and irradiance, the XGBoost model was used to predict relevant parameters of the energy storage system (battery). An 80 / 20 training / test set split was used, and a fixed random seed was employed to ensure reproducibility of the results. The hyperparameters of the XGBoost model are shown in Table 1.

[0144] Table 1

[0145]

[0146] The pre-trained energy storage start-up and shutdown state prediction model is used to predict the charging and discharging operation state of energy equipment, transforming the original mixed integer linear programming (MILP) problem into a linear programming (LP) problem. The results are compared with those obtained by the solver directly, as shown in Table 2.

[0147] Table 2

[0148]

[0149] Table 3 shows a comparison of the solution results between the two methods and the solution obtained directly from the solver:

[0150] Table 3

[0151]

[0152] Therefore, the capacity configuration determination method in the integrated energy system of this embodiment, while ensuring optimization accuracy, reduces the solution time for 8760 hours per year from 320 seconds to 3.5 seconds, saving approximately 98.9% of the calculation time, with a loss of only 1.9% in solution accuracy, significantly improving the solution efficiency in large-scale scenarios.

[0153] In this embodiment, a perfect policy set is used to train the energy storage start-up and shutdown state prediction model in the offline stage. In the online stage, the MILP solution is transformed into LP (linear programming) or low-dimensional constraint optimization through integer variable state prediction. This maintains the approximate optimality of the policy and significantly reduces the online computational complexity, greatly improving the solution efficiency. At the same time, the input of the energy storage start-up and shutdown state prediction model includes the equipment association parameters of energy equipment such as wind power, photovoltaic, load, and electricity price at future N time points, as well as the per-unit capacity vector of energy equipment. After training in multiple scenarios and configuration combinations, it has good generalization ability across seasons and energy equipment combinations, and is suitable for different operating conditions.

[0154] Example 2

[0155] In one specific embodiment, a system for determining capacity configuration in an integrated energy system is provided, such as... Figure 7 As shown, the determination system includes a construction module 100, a prediction module 200, a calculation module 300, an iteration module 400, and a determination module 500;

[0156] The construction module 100 is used to construct a population of candidate device capacities, including different combinations of first capacities, based on the device association parameters of multiple energy devices in the system to be constructed, using a genetic algorithm.

[0157] The prediction module 200 is used to input the time period to be predicted and the capacity population of alternative equipment into the energy storage start-up and shutdown state prediction model obtained by pre-training offline, and predict the actual energy storage start-up and shutdown operation state of each energy device under each target time boundary in different first capacity combinations.

[0158] Among them, the target time series boundary is obtained based on the period to be predicted, and the actual energy storage start-up and shutdown operation status is a characterization parameter for whether the energy equipment is in operation.

[0159] The calculation module 300 is used to calculate the fitness of different first capacity combinations based on the actual start-up and shutdown operation states corresponding to different first capacity combinations, using the fitness function of a genetic algorithm.

[0160] The iteration module 400 is used to iteratively update the candidate device capacity population based on fitness and return to call the prediction module 200 until the genetic algorithm meets the convergence condition or the fitness reaches the iteration stopping condition.

[0161] The determination module 500 is used to obtain the first capacity combination with the minimum fitness, which is used as the target capacity combination of the system to be built, thus obtaining the integrated energy system.

[0162] In one specific embodiment, the determining system further includes a pre-training module, which includes an acquisition unit, a solution unit, and a training unit;

[0163] The acquisition unit is used to acquire historical operation data and perform typical daily clustering processing on the historical operation data to construct a training sample set.

[0164] Among them, different typical days correspond to different sample time series boundaries;

[0165] The solution unit is used to solve the sample energy storage start-up and shutdown states of the energy devices in each sample capacity combination under different sample time boundaries, based on multiple pre-configured sample capacity combinations and training sample sets, using an optimization solver.

[0166] The training unit is used to train a preset model based on the training sample set, sample capacity combination and corresponding sample energy storage start-stop state, so as to obtain the energy storage start-stop state prediction model.

[0167] In one specific implementation, the solving unit is specifically used to normalize the configuration capacity of each energy device in each sample capacity combination to obtain the sample normalized capacity value of each energy device; to discretize the sample normalized capacity value based on a preset step size to obtain the sample discrete capacity value; and to use the sample discrete capacity value of each energy device in the sample capacity combination, the training sample set, and the candidate energy storage start-up and shutdown state as inputs to the optimization solver to solve the sample energy storage start-up and shutdown state of the energy device in each sample capacity combination under different sample time-series boundaries.

[0168] In this context, the energy storage start-stop status is represented by 0 to indicate that the energy equipment is not running, and 1 to indicate that the energy equipment is running.

[0169] In one specific implementation, the solving unit is implemented using the following formula:

[0170]

[0171] in, This indicates the start / stop status of the sample energy storage for each energy device. In the function The set of values ​​for the independent variable u when the minimum value is obtained, where C represents the sample size combination, u represents the start-stop state of the alternative energy storage, A and B represent weight coefficients, and d is the vector coefficient.

[0172] In one specific implementation, the training unit is specifically used to take the discrete capacity value of each energy device in the training sample set and sample capacity combination as the input of the preset model, and the corresponding sample energy storage start-up and shutdown state as the output of the preset model, to train the preset model and obtain the energy storage start-up and shutdown state prediction model.

[0173] In one specific implementation, the preset model includes the XGBoost model.

[0174] In one specific implementation, the prediction module 200 is specifically used to normalize the configured capacity of each energy device in the first capacity combination to obtain the first normalized capacity value of each energy device; to discretize the first normalized capacity value based on a preset step size to obtain the first discrete capacity value of the first capacity combination; and to input the time period to be predicted and the first discrete capacity value of the first capacity combination into the energy storage start-up and shutdown state prediction model to predict the actual energy storage start-up and shutdown operation state corresponding to different first capacity combinations.

[0175] In one specific implementation, the calculation module 300 is specifically used to calculate the operating cost of each first capacity combination during the predicted period in parallel based on the actual start-up and shutdown operation status of each first capacity combination; obtain the investment cost of each first capacity combination; and calculate the fitness of each first capacity combination based on the investment cost and operating cost using the fitness function of a genetic algorithm.

[0176] In one specific implementation, the iteration module 400 is specifically used to arrange different first capacity combinations in ascending order of fitness; a preset number of first capacity combinations are used as parent individuals before screening, and selection, crossover and mutation operations are performed on the parent individuals in sequence to generate a new generation of candidate device capacity population composed of new different first capacity combinations.

[0177] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs.

[0178] In this embodiment, a perfect policy set is used to train the energy storage start-up and shutdown state prediction model in the offline stage. In the online stage, the MILP solution is transformed into LP (linear programming) or low-dimensional constraint optimization through integer variable state prediction. This maintains the approximate optimality of the policy and significantly reduces the online computational complexity, greatly improving the solution efficiency. At the same time, the input of the energy storage start-up and shutdown state prediction model includes the equipment association parameters of energy equipment such as wind power, photovoltaic, load, and electricity price at future N time points, as well as the per-unit capacity vector of energy equipment. After training in multiple scenarios and configuration combinations, it has good generalization ability across seasons and energy equipment combinations, and is suitable for different operating conditions.

[0179] Example 3

[0180] Figure 8 This is a schematic diagram of the structure of an electronic device according to an example embodiment of the present disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the method for determining the capacity configuration in the integrated energy system described in any of the above embodiments. Figure 8 The electronic device 30 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0181] like Figure 8 As shown, the electronic device 30 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).

[0182] Bus 33 includes a data bus, an address bus, and a control bus.

[0183] The memory 32 may include volatile memory, such as random access memory (RAM) 321 and / or cache memory 322, and may further include read-only memory (ROM) 323.

[0184] The memory 32 may also include a program tool 325 (or utility) having a set (at least one) program module 324, such program module 324 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0185] The processor 31 executes various functional applications and data processing by running computer programs stored in the memory 32, such as the capacity configuration determination method in the integrated energy system provided in any of the above embodiments.

[0186] Electronic device 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed through input / output (I / O) interface 35. Furthermore, electronic device 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 36. As shown, network adapter 36 communicates with other modules of electronic device 30 via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0187] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0188] Example 4

[0189] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining capacity configuration in an integrated energy system provided in any of the above embodiments.

[0190] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0191] Example 5

[0192] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the method for determining capacity configuration in an integrated energy system provided in any of the above embodiments.

[0193] The program code for executing the computer program product of this disclosure can be written in any combination of one or more programming languages, and the program code can be executed entirely on a user device, partially on a user device, as a stand-alone software package, partially on a user device and partially on a remote device, or entirely on a remote device.

[0194] While specific embodiments of this disclosure have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of this disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of this disclosure, but all such changes and modifications fall within the scope of protection of this disclosure.

Claims

1. A method for determining capacity configuration in an integrated energy system, characterized in that, The determination method includes: S1. Based on the equipment association parameters of multiple energy devices in the system to be constructed, a genetic algorithm is used to construct a population of candidate equipment capacities including different combinations of first capacities. S2. Input the time period to be predicted and the capacity population of the candidate equipment into the energy storage start-up and shutdown state prediction model obtained by pre-offline training, and predict the actual energy storage start-up and shutdown operation state of each energy device under each target time boundary in different first capacity combinations. Wherein, the target time series boundary is obtained based on the time period to be predicted, and the actual energy storage start-stop operation status is a characterization parameter for whether the energy equipment is in operation. S3. Based on the actual energy storage start-up and shutdown operation states corresponding to different first capacity combinations, the fitness of different first capacity combinations is calculated using the fitness function of the genetic algorithm. S4. Iteratively update the candidate equipment capacity population according to the fitness, and return to step S2 until the genetic algorithm meets the convergence condition or the fitness reaches the iteration stop condition. S5. Obtain the first capacity combination with the smallest fitness as the target capacity combination of the system to be constructed, and obtain the integrated energy system.

2. The determination method according to claim 1, characterized in that, The training process of the energy storage start-up and shutdown state prediction model includes: S01. Obtain historical operation data, and perform typical day clustering processing on the historical operation data to construct a training sample set; Among them, different typical days correspond to different sample time series boundaries; S02. Based on the pre-configured multiple sample capacity combinations and the training sample set, an optimization solver is used to solve the energy storage start-up and shutdown states of the energy devices in each sample capacity combination under different sample time-series boundaries. S03. Based on the training sample set, the sample capacity combination, and the corresponding sample energy storage start-stop state, perform model training on the preset model to obtain the energy storage start-stop state prediction model.

3. The determination method according to claim 2, characterized in that, Step S02 includes: S021. Normalize the configuration capacity of each energy device in each sample capacity combination to obtain the sample normalized capacity value of each energy device; S022. Discretize the sample per-unit capacity value based on a preset step size to obtain the sample discrete capacity value. S023. Using the discrete capacity value of each energy device in the sample capacity combination, the training sample set, and the candidate energy storage start-stop state as input to the optimization solver, solve for the sample energy storage start-stop state of each energy device in the sample capacity combination under different sample time-series boundaries. In this context, the energy storage start / stop status is represented by 0 to indicate that the energy equipment is not running, and by 1 to indicate that the energy equipment is running.

4. The determination method according to claim 3, characterized in that, Step S023 is implemented using the following formula: in, This indicates the start / stop status of the sample energy storage corresponding to each of the aforementioned energy devices. In the function The set of values ​​of the independent variable u corresponding to the minimum value, where C represents the sample size combination, u represents the candidate energy storage start-stop state, A and B represent weight coefficients, and d is the vector coefficient; And / or, Step S03 includes: The discrete capacity value of each energy device in the training sample set and the sample capacity combination is used as the input of the preset model, and the corresponding sample energy storage start-up and shutdown state is used as the output of the preset model. The preset model is trained to obtain the energy storage start-up and shutdown state prediction model. And / or, The preset model includes the XGBoost model.

5. The determination method according to claim 1, characterized in that, Step S2 includes: S21. Normalize the configuration capacity of each energy device in the first capacity combination to obtain a first normalized capacity value for each energy device; S22. Discretize the first per-unit capacity value based on a preset step size to obtain the first discrete capacity value of the first capacity combination. S23. Input the first discrete capacity value of the time period to be predicted and the first capacity combination into the energy storage start-up and shutdown state prediction model to predict the actual energy storage start-up and shutdown operation state corresponding to different first capacity combinations.

6. The determination method according to claim 1, characterized in that, Step S3 includes: S31. Based on the actual start-up and shutdown operation status of each first capacity combination, calculate the operating cost of each first capacity combination in the predicted time period in parallel. S32. Obtain the investment cost for each of the first capacity combinations; S33. Using the fitness function of the genetic algorithm, the fitness of each first capacity combination is calculated based on the investment cost and the operating cost.

7. The determination method according to claim 1, characterized in that, Step S4 includes: S41. Arrange the different combinations of the first capacity in ascending order of fitness; S42. A preset number of the first capacity combinations are used as parent individuals. Selection, crossover, and mutation operations are performed on the parent individuals in sequence to generate a new generation of candidate device capacity population composed of new different first capacity combinations.

8. A system for determining capacity configuration in an integrated energy system, characterized in that, The determination system includes a construction module, a prediction module, a calculation module, an iteration module, and a determination module; The construction module is used to construct a population of candidate device capacities, including different combinations of first capacities, based on the device association parameters of multiple energy devices in the system to be constructed, using a genetic algorithm. The prediction module is used to input the time period to be predicted and the capacity population of the candidate equipment into the energy storage start-up and shutdown state prediction model obtained by pre-offline training, and to predict the actual energy storage start-up and shutdown operation state of each energy device under each target time boundary in different first capacity combinations. Wherein, the target time series boundary is obtained based on the time period to be predicted, and the actual energy storage start-stop operation status is a characterization parameter for whether the energy equipment is in operation. The calculation module is used to calculate the fitness of different first capacity combinations based on the actual energy storage start-up and shutdown operation states corresponding to different first capacity combinations, using the fitness function of the genetic algorithm. The iteration module is used to iteratively update the candidate equipment capacity population according to the fitness, and return to call the prediction module until the genetic algorithm meets the convergence condition or the fitness reaches the iteration stopping condition. The determining module is used to obtain the first capacity combination with the smallest fitness as the target capacity combination of the system to be built, thereby obtaining the integrated energy system.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the determination method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the determination method according to any one of claims 1 to 7.

Images