Management system and management device for a plurality of industrial machines supplied with power from a power supply facility
A management device includes a data acquisition unit that collects data related to power consumption supplied to a plurality of industrial machines from at least a power supply facility, a preprocessing unit that creates power consumption data based on the collected data, a decision making unit that determines a behavior of allocating a predetermined total suspension time to the plurality of industrial machines with respect to a current state of power consumption by the plurality of industrial machines with reference to a learning model associating a value of a behavior of allocating the predetermined total suspension time to the plurality of industrial machines with a state of power consumption by the plurality of industrial machines based on power consumption data, and a suspension time allocation unit that allocates the suspension time to the plurality of industrial machines.
1. A management device of a plurality of industrial machines supplied with power from a power supply facility, the management device comprising:
a processor configured to
collect data (i) related to power consumption supplied to the plurality of industrial machines from at least the power supply facility and (ii) obtained based on a detection value detected by a sensor, and
create power consumption data indicating information related to the power consumption supplied to the plurality of industrial machines during a predetermined time based on the collected data; and
a learning model storage for storing a learning model associating a value of a behavior of allocating a predetermined total suspension time to the plurality of industrial machines with a state of power consumption by the plurality of industrial machines, wherein
the processor is further configured to
determine a behavior of allocating the predetermined total suspension time to the plurality of industrial machines with respect to a current state of power consumption by the plurality of industrial machines using the learning model stored in the learning model storage based on the created power consumption data,
allocate the total suspension time to the plurality of industrial machines based on the determined behavior, and
decide the suspension time for each industrial machine to suspend an operation after a series of operations.
2. The management device according to claim 1 , wherein
the processor is further configured to
collect a suspension time after a series of operations of the industrial machine,
create the power consumption data, suspension time allocation data, and power consumption determination data, and
generate a learning model,
the suspension time allocation data indicate allocation of the predetermined total suspension time to the plurality of industrial machines after a state of power consumption indicated by the power consumption data is observed,
the power consumption determination data indicates a determination result for power consumption within a predetermined time by the plurality of industrial machines in the case where the total suspension time indicated by the power consumption data is allocated after the power consumption indicated by the power consumption data is observed, and
the learning model associates states of power consumption by the plurality of industrial machines with allocation behavior of the predetermined total suspension time to the plurality of industrial machines based on the power consumption data, the suspension time allocation data, and the power consumption determination data.
3. The management device according to claim 2 ,
wherein the processor is further configured to set a plus reward when average power consumed by the plurality of industrial machines within the predetermined time is reduced and set a minus reward when average power consumed by the plurality of industrial machines within the predetermined time is not reduced or when the average power increases, and generate the learning model based on a value of the reward.
4. The management device according to claim 1 , wherein the learning model is a behavior value table associating and storing a value of a behavior of allocating the predetermined total suspension time to the plurality of industrial machines with a state of power consumption by the plurality of industrial machines.
5. The management device according to claim 1 , wherein the learning model is a neural network having a multi-layer structure.
6. A management system, comprising:
a plurality of management devices connected to each other via a network, wherein
each management device among the plurality of management devices is for a plurality of industrial machines supplied with power from a power supply facility, and comprises:
a processor configured to
collect data (i) related to power consumption supplied to the plurality of industrial machines from at least the power supply facility and (ii) obtained based on a detection value detected by a sensor, and
create power consumption data indicating information related to the power consumption supplied to the plurality of industrial machines during a predetermined time based on the collected data; and
a learning model storage for storing a learning model associating a value of a behavior of allocating a predetermined total suspension time to the plurality of industrial machines with a state of power consumption by the plurality of industrial machines, wherein
the processor is further configured to
determine a behavior of allocating the predetermined total suspension time to the plurality of industrial machines with respect to a current state of power consumption by the plurality of industrial machines using the learning model stored in the learning model storage based on the created power consumption data,
allocate the total suspension time to the plurality of industrial machines based on the determined behavior, and
decide the suspension time for each industrial machine to suspend an operation after a series of operations,
the processor is further configured to
collect a suspension time after a series of operations of the industrial machine,
create the power consumption data, suspension time allocation data, and power consumption determination data, and
generate a learning model,
the suspension time allocation data indicate allocation of the predetermined total suspension time to the plurality of industrial machines after a state of power consumption indicated by the power consumption data is observed,
the power consumption determination data indicates a determination result for power consumption within a predetermined time by the plurality of industrial machines in the case where the total suspension time indicated by the power consumption data is allocated after the power consumption indicated by the power consumption data is observed,
the learning model associates states of power consumption by the plurality of industrial machines with allocation behavior of the predetermined total suspension time to the plurality of industrial machines based on the power consumption data, the suspension time allocation data, and the power consumption determination data, and
a learning result by the processor of one of the plurality of management devices is allowed to be shared among the plurality of management devices.