Short-term load forecasting method and forecasting system based on IPSO-chaotic BP network

Abstract

The present invention relates to a load short-term prediction method and prediction system based on IPSO-chaotic BP network, wherein the method first obtains actual historical power load data of a location to be tested, then inputs the power load data into the power load short-term prediction model of the IPSO-chaotic BP network, and obtains the power load prediction results of each hour of the prediction day of the location, wherein the power load short-term prediction model is based on the chaotic BP neural network, and the weights and thresholds in the chaotic BP neural network are obtained by improved particle swarm optimization, wherein the improved particle swarm adopts a nonlinear dynamic improved inertia weight strategy and an acceleration coefficient linear adjustment strategy that introduces an inverse sine function. The present invention can effectively improve the optimization speed and search performance of the algorithm, reduce the operation time of the algorithm, and improve the accuracy of short-term power load prediction, thereby reducing the operating cost of the power grid.

Description

Technical Field

[0001] The present invention relates to the field of power information technology processing, and in particular to a load short-term forecasting method and a forecasting system based on IPSO-chaotic BP network. Background Art

[0002] In recent years, the demand for electric energy has been increasing, and higher requirements have been placed on the quality of electric energy. Faced with increasing challenges, power system load forecasting has become an important part of power system planning and power equipment selection. Under the current circumstances, since electric energy cannot be stored in large quantities, the electric energy generated by the system must maintain a dynamic balance with the load changes of the system, otherwise it will affect the quality of power supply and even threaten the safe and stable operation of the system. If the load forecast value is too low, it will lead to tight power supply, unable to meet the needs of users, and greatly reduce the reliability of power supply; if the load forecast value is too high, it will lead to a decrease in the operating efficiency of power generation, transmission, and transformation equipment after they are put into the system, seriously affecting the economic indicators of the system. Therefore, more accurate load forecasting can reasonably arrange the start and stop of power plant units, ensure the safe and stable operation of the power system, reduce the occurrence of accidents, improve energy utilization, effectively reduce power generation costs, and improve economic and social benefits.

[0003] Electricity load forecasting can be divided into long-term, medium-term and short-term load forecasting. The results of long-term and medium-term load forecasting can determine the future construction of power plants and equipment installation, the size, location and time of unit capacity, the expansion and reconstruction of the power grid, and the construction and development of the power grid. Short-term load forecasting generally refers to the prediction of load conditions in the next day to a week. The purpose of short-term power load forecasting is to meet the accuracy requirements of demand forecast values ​​for power system production to the greatest extent. With the continuous improvement of the power market, low-error short-term power load forecasting will reduce waste.

[0004] The research on short-term load forecasting of power systems has a long history. Many experts and scholars at home and abroad have never stopped their research on load forecasting, and have made a lot of progress through continuous research. The more mature traditional methods include time series method, regression analysis method, trend extrapolation method, etc.; until the late 1980s, the expert system method was applied to short-term load forecasting of power systems, opening up a new field for forecasting methods; since the 1990s, artificial networks have received widespread attention from researchers in the application of power system load forecasting, and have continued to develop and achieved relatively satisfactory results; afterwards, with the continuous deepening of research, many load forecasting methods have been produced, such as: fuzzy theory, short-term load forecasting of power systems based on point pattern matching, and support vector machines. However, although there are many forecasting methods for power load, and each has its own advantages and disadvantages and different applicable occasions, the overall forecasting accuracy is not high. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to propose a load short-term forecasting method and forecasting system based on IPSO-chaotic BP network, which can effectively improve the optimization speed and search performance, reduce the calculation time, and improve the accuracy of short-term power load forecasting, thereby reducing the operating cost of the power grid.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] The present invention provides a method for short-term load forecasting based on IPSO-chaotic BP network, comprising the following steps:

[0008] Step S1: Obtaining actual historical power load data of the location to be tested;

[0009] Step S2: inputting the power load data into the power load short-term prediction model based on IPSO-chaotic BP network to obtain the power load prediction results at each hour of the prediction day of the location to be tested;

[0010] Among them, the short-term prediction model of power load is based on chaotic BP neural network, the weights and thresholds in the chaotic BP neural network are obtained by improved particle swarm optimization, and the improved particle swarm adopts nonlinear dynamic improved inertia weight strategy and acceleration coefficient linear adjustment strategy by introducing inverse sine function.

[0011] Preferably, in the nonlinear dynamic improved inertia weight strategy introducing the inverse sine function, the inertia weight ω(t) is expressed as:

[0012]

[0013] Where ω1 and ω2 are the inertia weights at the beginning and end, t represents the current number of iterations, and t max represents the maximum number of iterations of the population, γ is the adjustment factor, and rand() is a random value function between (0,1).

[0014] Preferably, the adjustment factor γ is in the range of 1.1 to 1.3.

[0015] Preferably, in the acceleration coefficient linear adjustment strategy, the adjustment formulas of the acceleration coefficients c1 and c2 are expressed as:

[0016]

[0017]

[0018] In the formula, c 11 、c 12 They represent the values ​​of the acceleration coefficient c1 at the beginning and end, respectively.21 , c 22 Respectively represent the starting and ending values ​​of the acceleration coefficient c2, N iter Represents the current number of iterations, and k is the correction constant.

[0019] Preferably, the correction constant k is in the range of 2.78 to 2.86.

[0020] The present invention also provides a short-term load forecasting system based on IPSO-chaotic BP network, comprising:

[0021] A historical data acquisition module is used to obtain actual historical power load data of the location to be tested;

[0022] An automatic prediction module, used for inputting the power load data into a short-term power load prediction model based on an IPSO-chaotic BP network, and obtaining power load prediction results at each hour of the prediction day for the location to be tested;

[0023] Among them, the short-term prediction model of power load is based on chaotic BP neural network, the weights and thresholds in the chaotic BP neural network are obtained by improved particle swarm optimization, and the improved particle swarm adopts nonlinear dynamic improved inertia weight strategy and acceleration coefficient linear adjustment strategy by introducing inverse sine function.

[0024] Preferably, in the nonlinear dynamic improved inertia weight strategy introducing the inverse sine function, the inertia weight ω(t) is expressed as:

[0025]

[0026] Where ω1 and ω2 are the inertia weights at the beginning and end, t represents the current number of iterations, and t max represents the maximum number of iterations of the population, γ is the adjustment factor, and rand() is a random value function between (0,1).

[0027] Preferably, in the acceleration coefficient linear adjustment strategy, the adjustment formulas of the acceleration coefficients c1 and c2 are expressed as:

[0028]

[0029]

[0030] In the formula, c 11 、c 12 They represent the values ​​of the acceleration coefficient c1 at the beginning and end, respectively. 21 , c 22 Respectively represent the starting and ending values ​​of the acceleration coefficient c2, N iter Represents the current number of iterations, and k is the correction constant.

[0031] The present invention also provides an electronic device, comprising:

[0032] One or more processors, a memory and one or more programs stored in the memory, wherein the one or more programs include instructions for executing the above-mentioned IPSO-chaotic BP network-based short-term load forecasting method.

[0033] The present invention also provides a computer-readable storage medium, comprising one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the above-mentioned IPSO-chaotic BP network-based short-term load forecasting method.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. The present invention proposes a chaotic BP network short-term prediction method based on improved particle swarm optimization, which not only overcomes the shortcomings of the standard particle swarm algorithm and the chaotic BP network algorithm in the iterative process, that is, the global search performance is poor and the real global optimal solution may not be found, but also maintains the advantages of the particle swarm algorithm such as simple parameter setting and easy operation, and also has satisfactory fitness;

[0036] 2. The present invention overcomes the shortcomings of the current algorithm such as local optimum and slow iterative convergence speed by improving the particle swarm optimization algorithm, effectively improves the optimization speed and search performance of the algorithm, and reduces the algorithm's operation time;

[0037] 3. The present invention effectively improves the accuracy of short-term power load forecasting, thereby reducing the operating cost of the power grid.

[0038] 4. The hybrid algorithm of the improved particle swarm optimization algorithm and the chaotic BP network proposed in the present invention solves the shortcomings of the traditional standard particle swarm algorithm, such as slow convergence speed and easy divergence, and also overcomes the shortcomings of the chaotic learning algorithm, such as low prediction accuracy and difficulty in data acquisition. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is the E of the three algorithms g Value comparison;

[0040] Figure 2 Comparison between the actual load curve and the load curves of the three algorithms;

[0041] Figure 3 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0042] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0043] This embodiment provides a method for short-term load forecasting based on IPSO-chaotic BP network, comprising the following steps:

[0044] Step S1: Obtaining actual historical power load data of the location to be tested;

[0045] Step S2: Input the power load data into the power load short-term prediction model based on IPSO-chaotic BP network to obtain the power load prediction results at each hour of the prediction day of the location. The power load short-term prediction model is based on the chaotic BP neural network. The weights and thresholds in the chaotic BP neural network are obtained by improved particle swarm optimization. The improved particle swarm adopts a nonlinear dynamic improved inertia weight strategy and an acceleration coefficient linear adjustment strategy that introduces an inverse sine function.

[0046] Specifically, Figure 3 As shown, the training steps of the power load forecasting model include:

[0047] Step S01: Initialize the particle swarm population information and the maximum allowed number of iterations T max The particle swarm population information includes the particle swarm population size, the initial position of the particle X i,j (0) and velocity and number of particles M;

[0048] Step S02: using an improved particle swarm optimization algorithm, by balancing global search and local search, respectively determining the global optimal position and local search position of each particle, and finding the weight and threshold of the chaotic BP network;

[0049] Step S03: According to the weight and threshold, the chaotic BP network is used to calculate the individual historical optimal fitness value corresponding to each particle, and the best individual historical optimal fitness value in the particle group is selected as the population historical optimal fitness value;

[0050] Step S04: Calculate the fitness value of each particle according to the fitness function of the particle;

[0051] Step S05: determine whether the calculated particle fitness value is greater than the individual historical optimal fitness value and the population historical optimal fitness value, if not, update the particle position and speed, if yes, return to step S02;

[0052] Step S06: Determine whether the process end condition is met. If so, stop iteration and output the result. If not, return to step S02. The process end condition is specifically that the difference between the hidden layer output value at the current moment and the hidden layer output value at the previous moment is less than the floating point relative accuracy.

[0053] The above training process can effectively improve the optimization speed and search performance of the algorithm, reduce the algorithm's operation time, and improve the accuracy of short-term power load forecasting, thereby reducing the operating cost of the power grid.

[0054] Specifically, in the improved particle swarm optimization algorithm, a nonlinear dynamic improved inertia weight strategy of the inverse sine function is introduced, and the inertia weight ω(t) is:

[0055]

[0056] Where ω1 and ω2 are the inertia weights at the beginning and end, t represents the current number of iterations, and t max Indicates the maximum number of iterations of the population; γ is the adjustment factor, which can be selected from 1.1 to 1.3, and rand() is a function that randomly takes values ​​between (0,1).

[0057] The improved inertia weight can ensure that the algorithm has good global optimization ability in the initial stage, can quickly find a range close to the global optimal solution, and has good local optimization performance at the end of the search, and can more reliably find the global optimal solution.

[0058] In the improved particle swarm optimization algorithm, the acceleration coefficients c1 and c2 are adjusted by using the acceleration coefficient linear adjustment strategy to ensure that particles can search the entire solution space as much as possible in the early stage of iterative search, and at the end of iterative search, the search particles are required to jump out of the local extreme value. The adjustment formula is expressed as:

[0059]

[0060]

[0061] In the formula, c 11 , c 12 They represent the values ​​of the acceleration coefficient C1 at the beginning and end, respectively. 21 , c 22 Respectively represent the starting and ending values ​​of the acceleration coefficient C2, N iter Indicates the current number of iterations; k is a correction constant, which can take a value of 2.78 to 2.86.

[0062] In addition to modifying the inertia weight ω(t) and acceleration coefficients c1 and c2 of the particle swarm algorithm itself, the improved particle swarm optimization algorithm is integrated with the chaotic BP neural network to complement each other, thereby achieving global optimization and improving the efficiency and solution accuracy of the algorithm. The output function x of the chaotic neuron in the chaotic BP network is j (t+1) is:

[0063]

[0064] Where t is the number of iterations, y i (t+1) is the internal state of the chaotic neuron.

[0065] The initial parameters of the model in this embodiment are: c1=1.516, c2=1.798, population number N=80, maximum number of iterations T max =5000. Set up a 3-layer network, with 3 layers of neurons in the input layer, 5 neurons in the hidden layer, and 1 neuron in the output layer. The output is the load output of the predicted day. In order to verify the practical application of the chaotic BP network model optimized by the improved particle swarm, the load data of a certain place was selected for testing and verification. First, the data was normalized, and the load data of the first 30 days were used as network training samples. The trained network was used as the verification model, and the load data of the last 30 days was used as the test data to verify the effect of the model in the short-term load forecasting process. The power load at 24h on a certain day was predicted using the standard particle swarm algorithm, the chaotic BP network, and the improved particle swarm optimized chaotic BP network algorithm proposed in the present invention. The comparison of evaluation indicators is shown in Table 1, and the load prediction results are shown in Table 2.

[0066] Table 1 Evaluation indicators E of three algorithms g Comparison Table

[0067]

[0068] As shown in Table 1, the chaotic BP network algorithm based on improved particle swarm optimization has a good performance in E g The value is smaller than the E of the standard particle swarm algorithm and the chaotic BP network algorithm after 3000 trainings. g The improved particle swarm optimization chaotic BP network algorithm has a good performance after 1500 training cycles. g The value is also smaller than the E when the standard particle swarm algorithm and the chaotic BP network algorithm are trained 5000 times. g The time used for the former is 75s, while the time used for the latter two is 373s and 306s respectively. It can be seen that the hybrid algorithm of improved particle swarm optimization adopted in this patent has obvious advantages in terms of convergence and training speed. The average percentage absolute error comparison between the prediction results of the three algorithms and the actual values ​​is as follows Figure 1 shown.

[0069] By comparing the prediction results and relative errors in Table 2, it can be seen that the load prediction accuracy of the chaotic BP network algorithm with improved particle swarm optimization when trained 400 times is significantly better than the prediction accuracy when the standard particle swarm algorithm (400 times) and the chaotic BP network algorithm (400 times), and the error fluctuation is small, indicating that the prediction method and model adopted by the present invention are significantly better than the other two algorithms in terms of prediction accuracy and speed.

[0070] Table 2 Comparison of prediction results of three algorithms

[0071]

[0072] refer to Figure 2 As shown in the figure, this embodiment uses the standard particle swarm algorithm, the chaotic BP network algorithm and the improved particle swarm optimized chaotic BP network algorithm to compare the actual load curve and the predicted load curve. As can be seen from the figure, the load curve predicted by the improved particle swarm optimized chaotic BP network algorithm is closer to the actual load curve, indicating that the prediction effect of the hybrid algorithm is better than that of the standard particle swarm algorithm and the chaotic BP network algorithm. Figure 1 and Figure 2 It can be seen that the relative error of the prediction results of the chaotic BP network algorithm with improved particle swarm optimization is controlled within 1.0%, and the error fluctuation is small.

[0073] The method of the present invention solves the shortcomings of slow convergence and easy divergence of the traditional standard particle swarm algorithm, and also overcomes the shortcomings of low prediction accuracy and difficulty in data acquisition of the chaotic learning algorithm. Different improved hybrid prediction models can be established according to the different daily load characteristics of working days and rest days. Practical application shows that compared with the traditional prediction method, the prediction method of the present invention is an effective load prediction method.

[0074] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0075] In another embodiment, an electronic device is provided, comprising one or more processors, a memory and one or more programs stored in the memory, wherein the one or more programs include instructions for executing the IPSO-chaotic BP network-based short-term load forecasting method.

[0076] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for short-term load forecasting based on IPSO-chaotic BP network, characterized in that: The steps include: Step S1: Obtaining actual historical power load data of the location to be tested; Step S2: inputting the power load data into the power load short-term prediction model based on IPSO-chaotic BP network to obtain the power load prediction results at each hour of the prediction day of the location to be tested; Among them, the short-term prediction model of power load is based on chaotic BP neural network, the weights and thresholds in the chaotic BP neural network are obtained by improved particle swarm optimization, and the improved particle swarm adopts nonlinear dynamic improved inertia weight strategy and acceleration coefficient linear adjustment strategy by introducing inverse sine function.

2. The method for short-term load forecasting based on IPSO-chaotic BP network according to claim 1 is characterized in that: In the nonlinear dynamic improved inertia weight strategy that introduces the inverse sine function, the inertia weight ω(t) is expressed as: Where ω1 and ω2 are the inertia weights at the beginning and end, t represents the current number of iterations, and t max represents the maximum number of iterations of the population, γ is the adjustment factor, and rand() is a random value function between (0,1).

3. The method for short-term load forecasting based on IPSO-chaotic BP network according to claim 2 is characterized in that: The adjustment factor γ ranges from 1.1 to 1.

3.

4. The method for short-term load forecasting based on IPSO-chaotic BP network according to claim 1 is characterized in that: In the acceleration coefficient linear adjustment strategy, the adjustment formulas of the acceleration coefficients c1 and c2 are expressed as: In the formula, c 11 、c 12 They represent the values ​​of the acceleration coefficient c1 at the beginning and end, respectively. 21 , c 22 Respectively represent the starting and ending values ​​of the acceleration coefficient c2, N iter Represents the current number of iterations, and k is the correction constant.

5. The method for short-term load forecasting based on IPSO-chaotic BP network according to claim 4 is characterized in that: The correction constant k ranges from 2.78 to 2.

86.

6. A short-term load forecasting system based on IPSO-chaotic BP network, characterized in that: include: A historical data acquisition module is used to obtain actual historical power load data of the location to be tested; An automatic prediction module, used for inputting the power load data into a short-term power load prediction model based on an IPSO-chaotic BP network, and obtaining power load prediction results at each hour of the prediction day for the location to be tested; Among them, the short-term prediction model of power load is based on chaotic BP neural network, the weights and thresholds in the chaotic BP neural network are obtained by improved particle swarm optimization, and the improved particle swarm adopts nonlinear dynamic improved inertia weight strategy and acceleration coefficient linear adjustment strategy by introducing inverse sine function.

7. The load short-term forecasting system based on IPSO-chaotic BP network according to claim 6 is characterized in that: In the nonlinear dynamic improved inertia weight strategy that introduces the inverse sine function, the inertia weight ω(t) is expressed as: Where ω1 and ω2 are the inertia weights at the beginning and end, respectively, t represents the current number of iterations, and t max represents the maximum number of iterations of the population, γ is the adjustment factor, and rand() is a random value function between (0,1).

8. The load short-term forecasting system based on IPSO-chaotic BP network according to claim 6 is characterized in that: In the acceleration coefficient linear adjustment strategy, the adjustment formulas of the acceleration coefficients c1 and c2 are expressed as: In the formula, c 11 、c 12 They represent the values ​​of the acceleration coefficient c1 at the beginning and end, respectively. 21 , c 22 Respectively represent the starting and ending values ​​of the acceleration coefficient c2, N iter Represents the current number of iterations, and k is the correction constant.

9. An electronic device, characterized in that: include: One or more processors, a memory and one or more programs stored in the memory, wherein the one or more programs include instructions for executing the short-term load forecasting method based on the IPSO-chaotic BP network as claimed in any one of claims 1-5.

10. A computer-readable storage medium, characterized in that: It comprises one or more programs executed by one or more processors of an electronic device, wherein the one or more programs comprise instructions for executing the short-term load forecasting method based on IPSO-chaotic BP network as claimed in any one of claims 1 to 5.

Images